Deepfakes and AI-Generated Content in Indian Elections: Rules, Risks, and Responses

Deepfakes and AI-generated content are now major risks in Indian elections because they can mislead voters through fake videos, cloned voices, manipulated images, and false political messages.

Indian campaigns must follow ECI guidance, MeitY rules, IT Rules, platform policies, and election law. AI can be used for safe tasks such as translation, subtitling, design, editing, and monitoring. But campaigns should not use AI to create fake speeches, false endorsements, cloned opponent voices, fake withdrawal claims, or synthetic content presented as real.

Deepfakes and AI-generated content have become a major concern in Indian election campaigning because they can make false political messages look real, spread quickly across social media, and damage voter trust before fact-checkers or authorities can respond. The 2024 Lok Sabha elections showed how fast synthetic media can enter political communication. AI-generated videos, cloned voices, manipulated images, translated speeches, parody clips, and fabricated endorsements all appeared in the campaign environment. Some uses were creative and harmless, such as language translation, visual editing, or support for campaign design. Others created serious risks, especially when synthetic media impersonated leaders, misrepresented speeches, or spread false claims during sensitive campaign periods.

The regulatory picture in India is still developing. There is no single deepfake election law that explains every permitted and prohibited use of AI in campaigns. Instead, campaigns must follow a combination of Election Commission of India directions, the Model Code of Conduct, the Representation of the People Act, the Information Technology Act, the Information Technology Rules, criminal law provisions, platform policies, and court orders. This creates a fragmented compliance environment for political parties, candidates, IT cells, consultants, and digital agencies. A campaign may use AI tools for production. Still, it cannot use them to deceive voters, impersonate a real person without clarity, spread false information, or hide the synthetic nature of campaign material when disclosure is expected.

The Election Commission of India acted during the 2024 Lok Sabha elections by warning political parties against the misuse of AI-based tools to create deepfakes that distort information or spread misinformation. On 6 May 2024, the Commission directed parties to avoid publishing or circulating deepfake audio and video, false or misleading information, derogatory content against women, content involving children in campaigns, and violent depictions involving animals. It also directed political parties to remove such content within three hours of it coming to their notice, warn the responsible person within the party, report unlawful information and fake accounts to the platforms, and escalate unresolved issues to the Grievance Appellate Committee under the IT Rules.

The ECI later moved toward stronger transparency standards for synthetic campaign material. In its 16 January 2025 advisory, the Commission advised political parties, leaders, candidates, and star campaigners to clearly label any image, video, audio, or other material generated or significantly altered using AI. The suggested labels include “AI-Generated,” “Digitally Enhanced,“ or “Synthetic Content.“ The advisory also asked parties to include disclaimers in campaign advertisements or promotional content that uses synthetic content. This is important because it does not ban every AI-generated campaign asset. Instead, it pushes campaigns toward transparency, especially when voters may otherwise mistake synthetic material for authentic footage, voice, or photography.

MeitY’s role is broader, as it regulates intermediaries and digital platforms under the Information Technology Act. The IT Rules, 2021 require intermediaries to follow due diligence obligations, operate grievance mechanisms, and respond to complaints about unlawful content. The updated version of the rules, as published by MeitY, covers digital media and intermediaries’ responsibilities under the Information Technology Act. These rules are not election-specific, but they matter during elections because deepfake content often spreads through intermediaries such as social media, video platforms, messaging services, and hosting services.

MeitY has also issued advisories to intermediaries about AI-generated content and unlawful content. The government’s position is that platforms should not allow their services to be used for content that violates Indian law, including content that spreads misinformation, impersonates individuals, or causes user harm. In October 2025, India proposed stricter rules requiring visible labeling of AI-generated media, metadata traceability, and user declarations for AI-generated uploads. Reuters reported that the draft proposal required AI-generated visual content to include markers covering at least 10 percent of the display area, and audio content to include an identifier during the initial 10 percent of the clip. This shows the direction of policy: India is moving from general takedown obligations toward visible labeling and traceability for synthetic media.

A deepfake in the election context is synthetic or manipulated media that falsely represents a person’s face, voice, speech, actions, or identity, potentially misleading voters. A realistic cloned voice of a candidate asking voters to support another party, a fabricated video showing a leader making a communal statement, or an AI-generated clip that makes a public figure appear to accept a bribe would fall into the high-risk category. The problem is not merely that AI was used. The problem is deception, impersonation, false attribution, and electoral harm.

Satire and parody sit in a more complex area. Indian election rules do not provide a clear, universal definition that distinguishes every deepfake from satire or political parody. A parody clip may be defensible when it is clearly humorous, visibly fictional, not presented as real footage, and not likely to mislead an ordinary voter. However, satire becomes risky when it uses realistic face swaps, cloned voices, fake news-style formatting, or fabricated statements without clear labeling. The safest campaign practice is to label synthetic or parody material prominently, avoid using real voices without permission, avoid realistic impersonation, and avoid any false claim about a candidate’s conduct, religion, caste, policy position, health, criminality, or private life.

For official campaign advertising, the use of AI-generated voices requires special caution. If a campaign uses an AI-generated voice that sounds like a real political leader, it should disclose that the voice is synthetic unless the voice is the leader’s real recorded voice and the campaign has proper authorization. If the AI voice imitates a leader, candidate, celebrity, journalist, citizen, or opponent, the legal risk increases sharply. It may raise issues of impersonation, misrepresentation, defamation, privacy, personality rights, platform policy violations, and election misinformation. The safer approach is to use original human voiceovers, licensed voice artists, clearly disclosed synthetic narration, or AI voices that do not imitate a real person.

AI tools for content production are not automatically illegal in Indian campaigns. A campaign can use AI for background design, translation assistance, subtitling, speech drafting, video editing, image resizing, data visualization, script variations, caption writing, and accessibility improvements. These uses are usually lower-risk when they do not fabricate facts, impersonate people, or conceal material alterations. The risk increases when AI generates a candidate’s speech, creates realistic images of events that never happened, modifies crowd sizes, fabricates endorsements, clones a leader’s voice, or creates attack content that appears to be documentary evidence. In Indian campaign practice, the compliance test should ask whether the AI asset could mislead voters about a real person, real event, real statement, or real endorsement.

Platforms such as Meta, YouTube, and X have their own policies on manipulated media, election misinformation, synthetic content, and political advertising. In practice, however, platform enforcement during the 2024 Indian election was uneven. The Guardian reported that Meta approved several AI-manipulated political advertisements submitted as a test by civil society groups, including ads that allegedly contained hate speech, disinformation, and violent messaging. According to the report, 14 out of 22 submitted ads were approved, and Meta’s systems failed to detect that the approved ads used AI-manipulated images. Meta said advertisers running political or election ads must follow authorization requirements and applicable laws, and that violating content can be removed regardless of how it was created.

This gap between policy and enforcement is one of the biggest risks for campaigns. A deepfake can spread for hours before takedown, and those hours may be enough to shape media coverage, WhatsApp circulation, local gossip, and voter perception. Platform action may depend on complaint quality, escalation channels, fact-checker involvement, media attention, legal notices, and whether the content clearly violates platform rules. This is why campaigns should not rely only on platforms. They need a first-response system that combines monitoring, evidence preservation, legal escalation, counter-messaging, journalist outreach, and voter-facing clarification.

The first six hours after identifying a deepfake are critical. The campaign should first preserve evidence before the content disappears or mutates. This means saving URLs, screenshots, screen recordings, timestamps, usernames, account IDs, captions, forwarding chains, and platform links. The team should also capture where the content appeared, how fast it is spreading, which language communities are affected, and whether party workers, influencers, or media pages are amplifying it. Evidence collection matters because platforms, police authorities, lawyers, fact-checkers, and election officials need clear, verifiable evidence.

The second step is verification. The campaign should not rush into a public denial before confirming whether the asset is fake, edited, old, taken out of context, AI-generated, or partly real. A digital forensics vendor, an internal video team, a trusted fact-checking partner, or a technical expert should inspect the media. They should check audio waveform patterns, lip sync errors, metadata, compression artifacts, source history, lighting inconsistencies, and whether the original footage exists. The campaign should prepare a short verification note that explains, in simple language, why the content is false or manipulated.

The third step is escalation. The campaign should report the content to the relevant platform via the election integrity, impersonation, manipulated media, harassment, defamation, or misinformation complaint routes. If the content violates election rules or appears during the Model Code of Conduct period, the campaign should also escalate it to the ECI or the relevant election authority. Since the ECI’s May 2024 direction expects parties to remove fake content within three hours of notice when it is under their control, parties should immediately take down any such content posted by their official handles, candidates, office-bearers, workers, or aligned pages.

The fourth step is public communication. The campaign should issue a clear denial from a verified handle, but the denial should not unnecessarily repost the deepfake in full quality. The better method is to use a blurred screenshot, a short warning label, and a direct statement that the content is fake or manipulated. The campaign should share the authentic version of the speech, event, or statement if available. It should also provide a simple instruction to supporters: do not forward the fake clip, report it on the platform, and share the verified clarification instead.

The fifth step is media and influencer response. A deepfake becomes more damaging when journalists, local pages, or political influencers discuss it without context. The campaign should send a short fact note to reporters, district-level media coordinators, spokespersons, booth-level WhatsApp admins, and friendly influencers. The note should include what is fake, the original fact, where the fake appeared, what action has been taken, and which verified post should be shared. The campaign should also prepare a spokesperson line for TV debates and digital interviews.

The sixth step is legal and operational follow-up. If the content is defamatory, impersonating, communal, obscene, threatening, or likely to disturb public order, the legal team should prepare complaints under the relevant election, IT, and criminal law provisions. The campaign should also map the origin and amplification path. If internal workers helped spread it, disciplinary action is necessary. If external networks are involved, the campaign should document repeat offenders and build a case file for regulators, platforms, and law enforcement.

For campaign content creators, the practical compliance rule is simple: AI may assist production, but it should not manufacture political reality. Using AI to clean audio, translate subtitles, generate neutral background graphics, resize creatives, write caption drafts, or edit non-deceptive videos is generally safer. Using AI to clone an opponent’s voice, create fake crowds, show fabricated meetings, simulate a leader saying something they never said, or produce inflammatory communal content is legally and reputationally dangerous. Even where the law is unsettled, voter deception remains the core risk.

A green, amber, and red approach can help campaign teams make faster decisions. Green use cases include AI-assisted grammar correction, subtitling, translation review, background music, non-realistic illustrations, layout design, image resizing, accessibility captions, and campaign analytics summaries. Amber use cases include AI-generated leader avatars, synthetic voiceovers, historical recreations, humorous parody clips, digitally enhanced rally visuals, and AI-translated speeches. These may be allowed only with consent, labeling, factual accuracy, and legal review. Red use cases include cloned voices of real leaders without disclosure, fake endorsements, fabricated scandals, fake news-style videos, manipulated communal content, fake polling claims, deepfake attack ads, and synthetic evidence of events that never occurred.

The blog should also explain that disclosure is becoming a central expectation. The ECI’s 2025 advisory specifically calls for prominent labeling of synthetic content in images, videos, audio, and other campaign materials. This means official campaign teams should build disclosure into their production workflow. Every AI-assisted video, synthetic voiceover, AI-generated image, or digitally altered political ad should be reviewed for whether a voter could mistake it for real media. If so, the content needs a visible label, audio disclosure, a caption note, or a disclaimer.

The key risk for Indian campaigns is not only legal punishment. It is a loss of credibility. A campaign that gets caught using deceptive AI can lose journalist trust, platform reach, voter confidence, and moral authority when it later complains about deepfakes targeting its own candidate. Responsible use of AI must therefore be treated as a campaign governance issue, not just a creative decision. Political digital heads should maintain an AI content register, an approval workflow, a legal review process, a source file archive, a consent record, and a crisis response protocol.

What Are the Current Rules for AI-Generated Political Content in Indian Elections?

India does not have a separate election law for AI-generated political content or deepfakes. You need to follow a combined set of rules drawn from the Election Commission of India, the Information Technology Act, the IT Rules, the Model Code of Conduct, criminal law, platform policies, and court directions.

For political campaigns, the main rule is simple: you can use AI for content production, but you cannot use it to mislead voters, impersonate people, distort facts, spread false claims, or publish synthetic media without clear labeling that prevents voters from mistaking it for real content.

The Election Commission of India has warned political parties against using AI tools to create deepfakes that distort information or spread misinformation. It also directed parties not to publish or circulate deepfake audio and video, false or misleading information, derogatory content against women, content using children in campaigns, or content showing violence or harm to animals.

Legal Review Note

You should treat this topic as a legal and campaign risk, not just a content issue. Before publishing AI-generated political ads, synthetic voiceovers, edited videos, leader avatars, or parody clips, your campaign should get a review from a lawyer familiar with Indian election law, IT law, media law, and platform policy.

Rules are still developing. Some areas remain unsettled, especially satire, parody, synthetic voices, digitally recreated leaders, and AI-translated speeches. When the law does not give a clear answer, use the safer test:

“Can an ordinary voter mistake this content for a real statement, real event, real endorsement, or real recording?”

If the answer is yes, label it clearly, keep consent records, and avoid using it as attack content.

What the Election Commission Says

The Election Commission’s 6 May 2024 direction is the main election-specific reference for AI misuse during the 2024 Lok Sabha elections. The ECI told political parties to avoid AI-based deepfakes that distort facts or spread misinformation. It also pointed parties to the Information Technology Act, the IT Rules, the Indian Penal Code, the Representation of the People Acts, and the Model Code of Conduct as the broader legal framework for misinformation and impersonation using deepfakes.

The ECI also placed a time-bound responsibility on political parties. If a deepfake or unlawful fake content comes to a party’s attention, the party must remove it within 3 hours if it is under the party’s control. The party must also warn the responsible person within the party, report fake accounts or unlawful information to the relevant platform, and escalate unresolved issues to the Grievance Appellate Committee under Rule 3A of the IT Rules.

For your campaign, this means you need a social media monitoring process. You cannot wait until the next day to review fake or manipulated content shared by your own workers, aligned pages, or local handles. You need a takedown workflow, a reporting workflow, and a record of every action taken.

AI Content Labeling Rules

The ECI issued a separate advisory on 16 January 2025 regarding labeling synthetic or AIAI-generated campaign content. It advised all political parties, leaders, candidates, and star campaigners to clearly label AI-generated or synthetic content shared through social media and other media platforms.

The advisory says parties should label any image, video, audio, or other material generated or significantly altered by AI. The ECI suggested clear labels such as:

• “AI Generated..”

• “Digitally Enhanc..ed.”

• “Synthetic Con..tent.”

The same advisory says campaign advertisements and promotional content should include disclaimers when synthetic content is used online or on other platforms.

For campaign teams, this creates a practical rule: do not publish AI-generated political content in a way that hides its synthetic nature. If you use AI to create or heavily alter a leader’s image, speech, audio, video, or rally visual, add a clear label.

What Counts as a Deepfake

A deepfake in the election context refers to synthetic or manipulated audio, video, images, or other visual content that makes a person appear to say, do, support, or endorse something they did not actually say, do, support, or endorse.

Common examples include:

• A cloned voice of a political leader asking voters to support another party

• A fake video showing a candidate making a communal or caste-based statement

• A synthetic clip showing a leader accepting money or making a private confession

• A fake endorsement by a celebrity, journalist, religious figure, or local influencer

• A manipulated speech that changes the meaning of the original statement

• A fake news-style video that presents AI-generated scenes as real footage

The ECI’s 2024 direction focuses on deepfake audio and video that distort information or spread misinformation. It also covers content that is false, untrue, or misleading.

Satire and Political Parody

Indian election rules do not provide a clear, single definition that distinguishes every deepfake from satire or political parody. That makes this area risky for campaigns.

You can reduce risk by applying these checks:

• Make the parody clearly fictional

• Do not use realistic cloned voices of real leaders without consent

• Do not use fake news formatting

• Do not show a real person making a fake factual claim

• Do not create confusion about a real event, real speech, or real endorsement

• Add a visible label when AI or synthetic media is used

• Avoid parody that targets religion, caste, women, children, or public order

A cartoon-style parody with a clear disclaimer carries lower risk. A realistic AI video of an opponent making a false statement carries high risk, even if the campaign later calls it satire.

Rules Under the IT Act and IT Rules

The Information Technology Rules, 2021, govern intermediaries, including social media platforms and other online services. The updated MeitY version states that the IT Rules came under the Information Technology Act, 2000, and replaced the earlier intermediary rules. It also defines audio, visual, or audiovisual information to include audio, images, photographs, graphics, video, moving visual recordings, sound recordings, and other content created, generated, modified, or altered through a computer resource.

This matters because AI-generated videos, edited images, cloned audio, synthetic graphics, and manipulated recordings fall within the broad digital content environment covered by intermediary rules.

MeitY also notified amendments to Rule 3(1)(d) of the IT Rules in 2025. These amendments address the removal of unlawful content by intermediaries and require more transparent, reasoned, and senior-level authorization for certain government takedown orders. The amended Rules came into effect from 15 November 2025.

For campaign managers, this means platforms have legal duties, but your campaign still needs its own evidence and escalation process. A weak complaint, lacking URLs, screenshots, timestamps, and legal grounds, slows action.

What Political Parties Must Avoid

Your campaign should not publish or amplify AI-generated political content that creates voter deception. The ECI specifically told parties to avoid deepfake audio and video, misinformation, false or misleading information, derogatory content toward women, the use of children in campaigns, and content showing violence or harm to animals.

Avoid these content types:

• Fake speeches by opponents

• AI-generated voice calls that impersonate real leaders

• Fake videos showing events that never happened

• AI-generated crowd visuals presented as real rally footage

• Fake endorsements by actors, influencers, caste leaders, religious leaders, or journalists

• Manipulated clips that change the meaning of a real speech

• Synthetic content that targets religion, caste, gender, or public order

• AI-generated “sting“ videos presented as evidence

• Fake screenshots of news reports, court orders, police notices, or ECI actions

If your campaign cannot prove the content is real, do not present it as fact.

Permissible AI Use in Campaign Content

AI is not banned in Indian election campaigning. You can use AI tools for routine content production, provided the output does not mislead voters or impersonate people.

Lower risk uses include:

• Grammar correction

• Caption writing

• Translation support

• Subtitle generation

• Video resizing

• Background cleanup

• Design variations

• Speech draft support

• Social media format adaptation

• Accessibility captions

• Basic image correction

• Data visualization

• Content scheduling support

These uses support production. They do not create fake political reality.

Higher risk uses need consent, labels, and legal review:

• AI-generated leader avatars

• Synthetic voiceovers

• AI translated speeches using a leader’s voice

• Digitally recreated historical leaders

• Satirical videos using real political figures

• AI-generated rally scenes

• AI edited attack ads

• AI-generated testimonial-style content

Do not publish these formats without clear disclosure and internal approval.

AI-Generated Voice of a Political Leader

An AI-generated voice of a political leader in an official campaign ad must be clearly disclosed when voters might mistake it for the leader’s real voice. The ECI’s 2025 advisory asks parties to label AI-generated or significantly altered audio, video, images, and other materials with clear notations such as “AI Generated,” “Digitally Enhanced,“ or “Synthetic Content.”

Use this practical rule:

“If the voice sounds like a real leader, disclose that it is synthetic unless it is an authorised real recording.”

The safest options are:

• Use the leader’s real recorded voice with permission

• Use a human voice artist who does not imitate a real person

• Use synthetic narration with a clear label

• Avoid cloning an opponent’s voice

• Keep written approval for official leader voice use

• Keep source files and scripts for audit

Voice cloning creates a serious risk because voters often trust audio messages in local WhatsApp groups, phone calls, and short videos.

Platform Rules and Enforcement Problems

Platforms have policies against manipulated media, election misinformation, hate speech, impersonation, and harmful political ads. But enforcement in the 2024 Indian election showed gaps.

The Guardian reported that Meta approved a series of AI-manipulated political advertisements submitted by civil society and accountability groups during India’s election. The report said researchers submitted 22 ads; 14 were approved, and Meta’s systems failed to detect that all approved ads used AI-manipulated images.

For your campaign, this means platform reporting alone is not enough. You need parallel action:

• Preserve evidence

• Report the content to the platform

• Send legal notices where needed

• Escalate to election authorities when election rules apply

• Publish a verified clarification

• Ask supporters not to forward the fake content

• Brief journalists and spokespersons

• Track repeat offenders

A fake clip can spread faster than a platform review queue. Your response system must act within hours.

What You Should Do When You Find a Deepfake

When your campaign discovers a deepfake targeting your candidate, act within the first 6 hours.

First, preserve proof. Save the URL, screenshots, screen recordings, post time, account name, handle, caption, comments, shares, and platform link. Capture the content before it gets deleted or edited.

Next, verify the content. Ask your media team, digital forensics partner, or legal team to check whether it is AI-generated, edited, old footage, mistranslated, or taken out of context.

Then report it. File complaints with the platform using the correct category, such as impersonation, manipulated media, election misinformation, defamation, hate speech, or harassment. If the content violates election rules, escalate it to the ECI or the relevant election authority.

After that, communicate clearly. Use your verified handles to flag content as fake or manipulated. Do not repost the full deepfake in high quality. Use a blurred screenshot, a clear warning, and the authentic clip or statement if available.

Finally, brief your network. Send a short fact note to spokespersons, district coordinators, WhatsApp admins, local media contacts, and influencers. Tell them what is fake, what the real fact is, and which verified post they should share.

Ways to Manage Deepfakes and AI-Generated Content in Indian Elections

Deepfakes and AI-generated content have become major concerns in Indian election campaigns. Political parties, campaign managers, IT cells, and digital teams now need clear systems to identify fake videos, cloned voices, manipulated images, and synthetic campaign material before they damage voter trust.

The safest approach is to use AI responsibly for content production while avoiding anything that misleads voters. Campaigns can use AI for translation, subtitles, design, editing, accessibility, and monitoring. But they should not use AI to create fake speeches, fake endorsements, cloned opponent voices, false withdrawal claims, or synthetic videos presented as real evidence.

Indian campaigns should also follow ECI guidance, MeitY rules, platform policies, and election compliance standards. Any AI-generated or significantly altered political content should carry a clear label, such as “AI Generated,” “Synthetic Content,“ or “Digitally Enhanced,“ to prevent voters from mistaking it for real content.

A strong campaign response should include early detection, evidence collection, platform reporting, legal escalation, public clarification, and local-language debunking. IT cells should monitor social media, WhatsApp alerts, Telegram channels, YouTube Shorts, Instagram Reels, and regional pages for suspicious content.

Way	Description
Detect Deepfakes Early	Monitor social media platforms, WhatsApp alerts, Telegram channels, YouTube Shorts, Instagram Reels, Facebook pages, X posts, and local news pages to identify fake videos, cloned voices, and manipulated visuals before they spread widely.
Verify Before Responding	Check whether the content is AI-generated, edited, dubbed, old footage, or taken out of context. Compare suspicious content with original speeches, event videos, public records, and verified campaign material.
Preserve Evidence	Save URLs, screenshots, screen recordings, account handles, timestamps, captions, comments, and engagement details. This helps with platform complaints, legal action, and election authority reporting.
Report to Platforms	Report deepfake content to Meta, YouTube, X, Instagram, Facebook, and other platforms under categories such as manipulated media, impersonation, misinformation, fake accounts, or election misinformation.
Escalate to Election Authorities	If the content misleads voters, impersonates a candidate, spreads false claims, or violates election rules, escalate it to the Election Commission or the relevant election authority.
Publish a Clear Clarification	Use verified campaign handles to state that the video, audio, or image is fake. Avoid reposting the full deepfake. Use a blurred screenshot or a warning label when needed.
Use AI Disclosure Labels	Label AI-generated or significantly altered content with clear terms such as “AI Generated,” “Synthetic Content,” “Digitally Enhanced,” or “AI Generated Voice” when voters can mistake it for real content.
Avoid Voice Cloning Misuse	Do not clone an opponent’s voice or make a leader appear to say something they did not say. If your campaign uses an AI-generated voice for your candidate, obtain consent, approval, and disclosure.
Separate Satire From Deception	Political parody and satire should be clearly fictional. Avoid realistic face swaps, cloned voices, fake news formats, fake documents, or fake statements that voters may treat as real.
Train IT Cells and Volunteers	Train digital teams, booth-level workers, social media admins, and volunteers to spot suspicious content, report it internally, and avoid forwarding fake videos or audio clips.
Create a Six-Hour Response Plan	Respond fast by saving proof, verifying the fake, reporting it, publishing a clarification, briefing journalists, alerting local teams, and tracking reuploads within the first six hours.
Maintain an AI Content Register	Keep records of AI tools used, prompts, source files, consent, disclosure labels, reviewer names, approval details, publication dates, and final content links.
Use AI for Safe Production Tasks	Use AI for grammar correction, translation, subtitles, poster design, video resizing, audio cleanup, accessibility captions, and issue monitoring. These uses are safer when reviewed by humans.
Avoid High-Risk AI Content	Do not create fake speeches, fake endorsements, fake voting instructions, fake withdrawal claims, fake ECI notices, fake police records, fake crowd visuals, or synthetic attack ads presented as real evidence.
Build Local-Language Debunks	Create clarifications in the same language and format for the fake content. Use short text, images, videos, and WhatsApp-safe messages for fast correction.
Monitor Reuploads	After reporting one fake post, continue tracking copies, cropped versions, dubbed versions, screenshots, memes, and reposts across platforms.
Use Legal Review for Sensitive Content	Get legal review for AI content involving caste, religion, women, children, voting process, violence, public order, criminal allegations, or opponent attacks.
Control Internal Sharing	Tell party workers and volunteers not to forward fake content. Ask them to share only the official clarification link and report the original post.
Follow Platform Policies	Follow the synthetic media, political ads, misinformation, impersonation, and manipulated media policies of Meta, YouTube, X, and other platforms.
Protect Voter Trust	Use AI responsibly to improve communication, not to mislead voters. The safest rule is: “Use AI to support production. Do not use AI to fake political reality.”

 

How Indian Political Campaigns Should Respond to Deepfake Attacks Within Six Hours

A deepfake attack can damage a political campaign before your team fully understands what happened. A fake video, cloned voice, edited speech, or synthetic image can move across WhatsApp groups, X, Facebook, Instagram, YouTube, Telegram, and local media pages within minutes. Your campaign cannot treat it as a normal social media issue. You need a fast, documented, legal, and public response.

The Election Commission of India has already warned political parties against using AI tools to create deepfakes that distort information or spread misinformation. The ECI also directed the parties to remove fake content within three hours of its coming to their notice, warn the responsible person within the party, report unlawful information and fake accounts to the platforms, and escalate unresolved issues to the Grievance Appellate Committee under Rule 3A of the IT Rules.

Your campaign should use a six-hour response system. The aim is simple: stop the spread, preserve the proof, verify the content, inform the public, report the violation, and protect the candidate’s credibility.

Why the First Six Hours Matter

The first six hours decide whether a deepfake remains a small digital incident or becomes a political crisis. Many voters do not wait for fact checks. They react to what they see first. A fake video can reach local WhatsApp groups, community pages, party workers, influencers, and media desks before your official denial appears.

Your campaign should respond fast, but not carelessly. A rushed denial without proof can create confusion, while a delayed response allows the fake content to set the public narrative.

Use this principle:

“Verify fast. Respond clearly. Do not amplify the fake content.”

You need one team to collect proof, one team to verify the media, one team to report it, one team to prepare the public response, and one senior person to approve the message.

Set Up a Deepfake Response Team Before the Attack

You cannot build a response system after a fake video goes viral. You need the team before campaign season starts.

Your deepfake response team should include:

• Digital monitoring lead

• Legal advisor

• Social media manager

• Video editor or forensic reviewer

• Candidate office representative

• Media relations lead

• IT cell coordinator

• War room decision maker

• Platform escalation contact

• Local language content lead

Each person needs a clear role. Do not let five people draft five different responses. Assign one approval path and one public message.

Your campaign should also maintain:

• Verified social media handles list

• Official spokesperson list

• Platform reporting links

• Journalist contact list

• District WhatsApp admin list

• Legal complaint format

• Candidate photo, video, and speech archive

• Original rally footage archive

• Media monitoring dashboard

• Emergency response templates

This preparation saves time when a fake clip appears.

Identify the Deepfake and Freeze the Evidence

The first action is not posting a denial. The first action is saving proof.

Deepfake posts can disappear, get edited, move to new accounts, or appear in cropped versions. Your team must preserve the original evidence before it changes.

Save the following:

• Full URL of the post

• Screenshot of the post

• Screen recording of the video or audio

• Date and time of discovery

• Account name and handle

• Profile URL

• Caption text

• Hashtags or keywords used

• Comments showing early spread

• Share count, view count, and engagement count

• Platform name

• WhatsApp forward screenshot, if available

• Telegram channel link, if available

• Names of pages or influencers spreading it

• Any paid ad identifier, if visible

Do not rely on memory. Save files in a single folder, with the date, time, platform, and content type.

Use this naming format:

“2026-05-19_X_FakeVideo_CandidateName_FirstSource.mp4”

This helps your legal team, platform team, and media team work from the same record.

Verify Whether the Content Is Fake, Edited, or Taken Out of Context

Not every harmful clip is a deepfake. Some clips are old videos shared as new. Some are edited speeches. Some are mistranslations. Some are parody clips presented as real. Some are AI-generated.

Your campaign must classify the content before it responds.

Check these points:

• Does the face movement match the voice?

• Does the lip movement look unnatural?

• Does the audio tone shift between words?

• Does the video show strange lighting near the mouth or eyes?

• Does the background match the claimed location?

• Does the candidate’s clothing match known event footage?

• Does the clip cut before or after a key statement?

• Does the same clip exist in old news archives?

• Does the audio match another speech?

• Does the content include fake subtitles?

• Does the post use a fake news template?

• Does the content claim a false date, place, or event?

Your media team should compare the clip with original speeches, rally videos, press conferences, and official recordings. If you have a forensic partner, send the file immediately. If not, your internal video team should still complete a first-level review.

Do not say “AI generated” unless you have a reasonable basis. Use careful language when the technical proof is still under review.

Better wording:

“This video is fake and manipulated. The candidate did not make this statement.”

Avoid weak wording:

“This may be fake.”

When you know the statement is false, say it directly.

Classify the Risk Level

Not every deepfake requires the same response. A small parody clip on a low-reach page needs to be monitored and reported. A fake communal speech during voting week needs immediate legal, platform, and media escalation.

Classify the attack by risk:

Low Risk

• Obvious parody

• Low reach

• No realistic impersonation

• No harmful claim

• No confusion among voters

Response: monitor, archive, report if needed, avoid overreaction.

Medium Risk

• Edited speech

• Fake quote card

• Misleading caption

• Local page amplification

• Confusion among party workers or supporters

Response: issue clarification, the platform report, and brief local teams.

High Risk

• Realistic cloned voice

• Fake video of candidate speech

• Communal or caste-based content

• Fake criminal allegation

• Fake withdrawal or alliance claim

• Fake endorsement

• Spread by verified accounts or media pages

• Viral WhatsApp circulation

Response: public denial, platform escalation, legal complaint, ECI escalation, media briefing, district-level counter-distribution.

Report the Content to Platforms

Your team should report the fake content to every platform where it appears. Use the strongest relevant category. Do not submit a generic complaint if the platform offers specific options.

Use categories such as:

• Impersonation

• Manipulated media

• Election misinformation

• Defamation

• Harassment

• Hate speech

• Violence or public order risk

• Fake account

• Unauthorized use of image or voice

• Synthetic media without disclosure

Include clear proof in the complaint:

• Link to the fake post

• Screenshot

• Screen recording

• Explanation of why it is fake

• Link to the original video or speech, if available

• Candidate’s verified profile link

• Legal or election rule reference

• Request for urgent removal or visibility reduction

Platform enforcement during Indian elections has shown gaps. The Guardian reported in May 2024 that Meta approved several AI-manipulated political ads submitted by researchers as a test, including those that used AI-generated images. The report said 14 out of 22 submitted ads were approved.

This means your campaign should not rely solely on platform action. Report the content, but continue with public clarification, media response, and legal escalation.

Escalate to Election Authorities and Legal Channels

If the deepfake violates election rules, your campaign should escalate it to the proper authority. The ECI’s 6 May 2024 direction specifically told parties to refrain from publishing or circulating deepfake audio and video, misinformation, false or misleading information, derogatory content toward women, content involving children in campaigns, and depictions of violence or harm to animals.

Escalate the issue when the content:

• Misleads voters

• Impersonates a candidate

• Uses cloned voice or fake video

• Creates communal or caste tension

• Defames a candidate

• Claims fake withdrawal from the election

• Claims fake alliance or endorsement

• Targets women with derogatory content

• Uses children in campaign content

• Spreads during the silence period or the voting period

• Uses fake official notices or fake ECI material

Your legal complaint should include:

• Summary of the incident

• Candidate name and constituency

• Platform links

• Screenshots and recordings

• Time of first detection

• Accounts spreading the content

• Why the content is false or manipulated

• Election harm caused

• Requested action

• Contact person for follow-up

Keep the complaint short and factual. Do not add political commentary inside the legal complaint.

Issue a Public Clarification From Verified Handles

Your public clarification should come from verified and trusted sources. Use the candidate’s official handle, party handle, spokesperson handle, and campaign website.

Keep the statement short.

Example:

“A fake and manipulated video is being circulated in my name. I have not made this statement. Please do not forward it. We have reported it to the platform and election authorities. Follow only my verified handles for official updates.”

For a cloned voice message:

“An AI-generated voice message is being circulated as my voice. This audio is fake. I have not issued this message. Please report it and do not forward it.”

For a fake withdrawal claim:

“The claim that I have withdrawn from the election is false. I remain the official candidate. Please ignore fake posts and follow verified updates only.”

Do not repost the full fake video in high quality. That helps the attacker. Use a blurred screenshot or a short-cropped frame with a warning label.

Use clear labels:

• “Fake Video..”

• “Manipulated Aud..io”

• “False C..laim.”

• “Do Not F..orward.”

• “Report This.. Content”

Share the Original Proof When Available

A denial works better when you show the real source. If the deepfake uses edited speech footage, post the full original clip. If it uses fake subtitles, post the original speech with correct subtitles. If it uses a fake location claim, -ntpost-verified event, on-the-fly footage, or official schedule details.

Your clarification should answer three questions:

• What is fake?

• What is the real fact?

• What should people do now?

Example:

“The viral clip cuts and changes the meaning of my speech. The full speech is available here. The viral version is manipulated. Please share this verified clip instead.”

This helps supporters, journalists, and fact-checkers correct the record.

Brief Journalists, Influencers, and Local Campaign Teams

Deepfakes often spread through local pages and informal networks before the news media reports them. Your response must reach the same channels.

Send a short fact note to:

• Political reporters

• Local journalists

• District media coordinators

• Party spokespersons

• Booth-level WhatsApp admins

• Influencers who support the campaign

• Community page admins

• Constituency volunteers

• Legal and compliance teams

• Candidate office staff

Your fact note should include:

• One-line issue summary

• What is fake

• What is true

• Link to verified clarification

• Link to original footage, if available

• Instruction not to forward the fake content

• Platform reporting instruction

• Approved spokesperson line

Do not send long explanations to local teams. Give them a clean message they can copy accurately.

Control Internal Sharing

Many campaigns accidentally spread the deepfake while trying to expose it. Party workers forward the fake clip to groups with messages like “See what they are spreading.” This increases reach.

Give your workers a strict instruction:

“Do not forward the fake video. Share only the official clarification link.”

Your internal message should say:

• Do not download and repost the fake clip

• Do not add angry captions

• Do not tag journalists with the fake video

• Do not share unverified forensic claims

• Do not attack communities or voters

• Share only the approved clarification

• Report the original post through platform tools

This step reduces accidental amplification.

Use a Six-Hour Response Protocol

First 30 Minutes

Your monitoring team identifies the content, saves the link, records the post, captures screenshots, and alerts the response group. The team checks whether the content appears on multiple platforms.

Action list:

• Save original link

• Take screenshots

• Record the video or audio

• Capture account details

• Alert legal, media, and social teams

• Create incident folder

• Stop internal forwarding of the fake content

30 to 60 Minutes

Your verification team reviews the content. The candidate’s office confirms whether the statement, event, or claim is false. The media team searches for original footage.

Action list:

• Compare with original speech or event

• Check source account history

• Check whether the content is edited, old, or synthetic

• Identify affected language and region

• Classify risk level

• Draft first internal fact note

One to Two Hours

Your team reports the content to platforms and prepares the public clarification. The legal team drafts the complaint if the content poses an election, defamation, communal, or public order risk.

Action list:

• Report to the platform

• Prepare legal complaint

• Draft verified clarification

• Prepare blurred screenshot, if needed

• Prepare spokesperson line

• Prepare WhatsApp-safe clarification text

Two to Three Hours

Your campaign publishes the official clarification and sends the fact note to internal teams. If the content came from a party worker or an aligned page, remove it immediately. ECI’s 2024 direction requires parties to remove fake content within three hours of becoming aware of it and when it is under their control.

Action list:

• Publish verified clarification

• Share official link with district teams

• Ask workers not to forward the fake clip

• Remove content from party-controlled pages

• Warn responsible internal persons

• Report fake accounts to platforms

Three to Four Hours

Your media team contacts journalists, fact-checkers, and influencers. The aim is to stop wrong reporting and push verified information.

Action list:

• Send media note

• Send original proof, if available

• Offer spokesperson comment

• Request correction from the pages sharing the fake

• Ask influencers to share the official clarification

• Track new versions of the fake content

Four to Six Hours

Your team reviews the spread, checks the platform action, escalates unresolved complaints, and prepares the next response if the fake content continues to spread.

Action list:

• Track removals and non-removals

• Record platform complaint IDs

• Escalate unresolved posts

• Send legal notices where needed

• Update media and local teams

• Prepare a second clarification if needed

• Document the incident for post-crisis review

Prepare Approved Response Templates

Your campaign should prepare templates before the election season. Speed matters, but accuracy matters too.

Fake Video Template

“A fake and manipulated video is being circulated about [candidate name]. The statement shown in the video is false. [Candidate name] did not say this. We have reported the content to the platform and authorities. Please do not forward it. Follow only verified handles for official updates.”

Fake Audio Template

“An AI-generated audio clip is being circulated as [candidate name] ‘ss voice. This audio is fake. [Candidate name] has not issued this message. Please report the clip and share only this official clarification.”

Fake Withdrawal Template. The claim that [candidate name] has withdrawn from the election is false. [Candidate name] remains the official candidate. Please ignore fake posts and follow verified campaign channels.”

Fake Endorsement Template. A fake endorsement post is circulating under the name of [person or group]. This claim is false. No such endorsement has been made. We have reported the content for removal.”

Do Not Make These Mistakes

Campaign teams often make deepfake attacks worse by reacting emotionally.

Avoid these mistakes:

• Reposting the full deepfake

• Calling it fake without checking facts

• Sharing unverified technical claims

• Using abusive language

• Blaming a community without proof

• Waiting for platform action before public clarification

• Sending different explanations from different handles

• Allowing party workers to forward the fake clip

• Ignoring local language versions

• Deleting evidence before saving it

• Failing to brief journalists

• Treating WhatsApp as a secondary channel

Your response should stay disciplined. One message. One proof line. One action instruction.

Build a Deepfake Monitoring System

A campaign needs daily monitoring during election season. Deepfake attacks rarely start on national media. They start in small pages, anonymous accounts, meme handles, Telegram channels, WhatsApp groups, and local video pages.

Monitor these areas:

• Candidate name in English and regional languages

• Nicknames and common misspellings

• Constituency keywords

• Opponent keywords

• Fake resignation or withdrawal claims

• Fake alliance claims

• Fake scandal claims

• Voice note circulation

• Short video platforms

• Local news pages

• Meme pages

• Telegram channels

• WhatsApp reports from booth teams

Train local workers to report suspicious content using a single internal form. Ask them to submit links, screenshots, and group details instead of forwarding the fake content.

Use Labeling and Transparency in Your Own AI Content

Your campaign loses credibility if it attacks deepfakes while using deceptive AI content itself. The ECI’s 2025 advisory asked political parties, candidates, and campaigners to label AI-generated or synthetic campaign content and use notations such as”AI Generated”, “Digitally Enhanced“, or “Synthetic Content.”

Apply the same rule internally.

If your campaign uses AI-generated video, synthetic voice, altered images, or digital recreations, clearly label them. Keep consent records and source files. Do not use AI to create fake events, fake speeches, fake endorsements, or fake evidence.

Use this rule: “Do not demand clean politics from opponents while publishing unclear synthetic content yourself.”

Keep Records for Legal and Election Review

Every deepfake incident should become a case file. Do not treat it as a one-time social media problem.

Your case file should include:

• Original content copy

• Screenshots

• URLs

• Account details

• Platform complaint IDs

• Legal complaints

• ECI complaint copy, if filed

• Public clarification copy

• Media note

• Internal instructions

• Timeline of action

• List of accounts that amplified the content

• outcome

• Lessons for the next incident

This record helps if the content returns, if journalists ask for details, or if authorities need proof.

What the ECI Advisory Said About AI-Generated Content in the 2024 Elections

The Election Commission of India addressed AI-generated content during the 2024 Lok Sabha elections because deepfake audio, fake videos, edited speeches, impersonation content, and false campaign material began spreading across social media platforms. The ECI did not issue a separate AI law in 2024. Instead, it reminded political parties that existing election law, IT law, criminal law, and the Model Code of Conduct already apply to synthetic and manipulated content.

On 6 May 2024, the ECI issued directions for the responsible and ethical use of social media by political parties and their representatives. The advisory warned parties against using AI-based tools to create deepfakes that distort information or spread misinformation. It also brought the existing legal framework for misinformation and synthetic content to the notice of political parties.

The advisory gave campaign teams a clear message: “Do not use AI to fake political reality, impersonate people, or mislead voters.”

For you as a campaign manager, this means AI can support production work. Still, it cannot create false speeches, fake endorsements, manipulated videos, cloned voices, or misleading visuals that voters may treat as real.

Why the ECI Issued the Advisory

The 2024 Lok Sabha elections became the first major Indian national election in which AI-generated political content appeared on a visible scale. Campaigns, supporters, influencers, and anonymous pages used AI tools to create videos, memes, edited speeches, voice content, and attack material. Some content supported harmless communication. Other content created voter confusion.

The ECI acted because deepfakes create three direct risks:

• They can impersonate candidates or leaders

• They can spread false claims before fact checks reach voters

• They can damage voter trust during a sensitive election period

The ECI advisory focused on political parties because party workers, social media teams, local pages, and candidate networks often control the first layer of campaign distribution. If these networks circulate fake content, the damage spreads quickly.

What the Advisory Told Political Parties Not to Do

The ECI asked political parties to avoid publishing or circulating content that violates election law, social media ethics, and public order. It specifically warned against AI-based deepfakes and misleading digital content.

The advisory covered these content types:

• Deepfake audio and video

• False or misleading information

• Misinformation and impersonation using deepfakes

• Content that distorts facts

• Derogatory content against women

• Content involving children in political campaigning

• Content showing violence or harm to animals

• Unlawful information shared through fake accounts

• Social media content that violates the Model Code of Conduct

The ECI direction matters because it treats synthetic content as part of the wider misinformation problem. A campaign cannot defend fake content by saying “AI made it ” or “it was only digital editing.” If the content misleads voters, impersonates someone, or harms election integrity, the campaign faces risk.

The Three-Hour Removal Requirement

The strongest operational instruction in the 2024 advisory was the three-hour removal rule. The ECI directed political parties to remove fake content within three hours after it comes to their notice, when the content is under their control.

This applies to content shared by:

• Official party handles

• Candidate pages

• Party workers

• Campaign volunteers

• District pages

• Local social media teams

• Party-controlled WhatsApp or Telegram networks

• Campaign-linked influencer pages, where control or coordination exists

For your campaign, this creates a clear duty. Once your team finds fake or manipulated content within your network, you must act quickly. Delete it. Warn the person who posted it. Save proof of removal. Keep records.

Do not wait for a legal notice. Do not wait for media coverage. Do not let local teams continue circulating it.

Warning:g Responsible Party Workers

The ECI also said political parties must warn the responsible person within the party when fake content comes to notice. This means campaigns need internal discipline, not just public denial.

Your campaign should create a written internal rule: “No worker, volunteer, page admin, or influencer linked to the campaign should share AI-generated or manipulated political content without approval.”

If someone shares fake content, your campaign should:

• Ask them to remove it

• Send a written warning

• Record the warning

• Remove repeat offenders from campaign groups

• Review who approved or forwarded the content

• Train local teams again

This protects the campaign from repeat incidents. It also shows authorities that the party acted after discovering the violation.

Reporting Fake Accounts and Unlawful Content to Platforms

The advisory also asked parties to report fake accounts and unlawful information to the relevant social media platforms.

This matters because deepfake attacks often come from anonymous pages, fake profiles, proxy handles, meme accounts, or newly created local pages. Your campaign should not only deny the content but also address the underlying issues. You should file platform complaints with proof.

Your platform complaint should include:

• Link to the fake post

• Screenshot of the content

• Screen recording, whether it is video or audio

• Account name and handle

• Date and time of discovery

• Reason the content is fake or manipulated

• Link to the real speech or source material, if available

• Request for removal under manipulated media, impersonation, misinformation, or election rules

A weak complaint slows action. A complete complaint improves your chance of removal.

Escalation to the Grievance Appellate Committee

The ECI advisory told parties to escalate unresolved issues to the Grievance Appellate Committee under Rule 3A of the IT Rules.

This gives campaigns another route when platforms do not act on complaints. If your team reports a deepfake and the platform does not remove it, your legal team should prepare an escalation file.

The file should include:

• Original complaint copy

• Platform response, if any

• Complaint reference number

• URLs and screenshots

• Evidence that the content is fake

• Explanation of election harm

• Candidate or party authorization letter

• Request for urgent action

You need proper documentation. Do not rely on screenshots; if you can, also save URLs, timestamps, and complaint IDs.

Legal Framework Mentioned by the ECI

The ECI did not create a new legal category only for AI content in the 2024 advisory. Instead, it pointed to existing laws and rules addressing misinformation, impersonation, unlawful content, and election conduct.

The ECI referred to the existing legal framework governing misinformation and synthetic content, including the Information Technology Act, the IT Rules, the Indian Penal Code, the Representation of the People Acts, and the Model Code of Conduct.

For campaign teams, this means AI-generated political content can create several types of risk at once:

• Election law risk

• Criminal law risk

• IT law risk

• Defamation risk

• Platform policy risk

• Model Code of Conduct risk

• Candidate reputation risk

Do not treat AI content solely as a technical issue. Treat it as a legal, political, and communications risk.

What the Advisory Means for Campaign Content Teams

The 2024 advisory changes how campaign teams should approve content. Every official digital unit should check whether a post uses AI, editing, synthetic media, cloned voice, or fake-news-style formatting.

Before publishing content, your team should ask:

• Does this content show a real leader saying something they did not say?

• Does it use a cloned or synthetic voice?

• Does it show an event that did not happen?

• Does it change the meaning of a real speech?

• Does it create a fake endorsement?

• Does it use fake news graphics or fake official documents?

• Does it target religion, caste, gender, or public order?

• Does it involve children in campaign material?

• Does it include derogatory content against women?

• Can an ordinary voter mistake it for real content?

If the answer raises concern, do not publish it without legal and senior campaign approval.

What the Advisory Means for IT Cells

The advisory affects IT cells, tent production, volunteer networks, social listening, meme pages, short videos, WhatsApp distribution, and influencer coordination.

Your IT cell should create these internal controls:

• A pre-publication review process of AI-generated content

• A separate approval path for videos, voices, and leader images

• A takedown process for fake content shared by internal teams

• A reporting process for fake accounts and deepfake attacks

• A content archive with source files and approvals

• A daily monitoring system for candidate name, party name, and constituency keywords

• A legal escalation process for high-risk fake content

• A training note for local admins and volunteers

The advisory makes one thing clear: if your network circulates fake content, your campaign cannot say the main office did not know. You need internal monitoring.

What the Advisory Means for Deepfake Attacks Against Your Candidate

The advisory also helps campaigns respond when they become targets. If a fake video or cloned voice attacks your candidate, you can use the ECI direction as part of your complaint.

Your response should include:

• Save the original link and screenshots

• Record the fake video or audio

• Verify whether it is edited, synthetic, old, or taken out of context

• Report it to the platform

• Escalate the  issue to the election authorities if it affects voters

• Publish a verified clarification

• Ask supporters not to forward the fake content

• Share the official clarification link

• Brief journalists and local campaign teams

• Track whether the platform removes it

Use this response line: “This content is fake and manipulated. It misleads voters. We have reported it to the platform and election authorities.”

Avoid reposting the full fake video. Use a blurred screenshot or warning graphic.

What the 2024 Advisory Did Not Fully Settle

The 2024 issue gave clear warnings, but it did not answer every question. Some areas still need careful legal review.

Unsettled areas include:

• Satire usAI-generated leader voices

• Political parody with realistic face swaps

• AI translated speeches using a leader’s voice

• Synthetic avatars of living politAI-generated historical recreations

• Digitally enhanced crowd vAI-generated memes that resemble real evidence

• Synthetic campaign ads that use real leader likeness

The safer rule is direct: “If voters can mistake synthetic content for real content, label it clearly or do not use it.”

How the 2025 ECI Advisory Built on the 2024 Position

The 2024 advisory focuses on the responsible use of social media, the removal of fake content, the reporting of fake accounts, and action against deepfakes. In January 2025, the ECI moved further by advising parties to use AI-generated and synthetic campaign content.

The 16 January 2025 advisory asked political parties, candidates, and campaigners to clearly label syntAI-generated material used in election campaigning. It referred to labels such as”AI Generated, “”Digitally Enhanced,” and “Synthetic Content.”

This later advisory strengthens the compliance direction. It tells campaigns that transparency matters when they use AI. It does not ban every AI tool. It asks parties to inform voters when content is synthetic or has been significantly altered.

Practical Compliance Rule for Campaigns

Your campaign should follow a simple rule: AI-generated political content: “Use AI for production support. Do not use AI for deception.”

Lower risk AI uses include:

• Grammar correction

• Subtitle generation

• Translation support

• Video resizing

• Design variation

• Audio cleanup

• Accessibility captions

• Social media format adaptation

• Background correction

• Drafting campaign High-risk

High-risk AI uses include:

• Cloning a leader’s voice

• Creating fake speeches

• Showing fake events

• Creating fake crowd visuals

• Producing fake endorsements

• Altering an opponent’s speech

• Creating fake sting videos

• Creating news-style attack videos

• Using synthetic content without clear labels

Your content team should help users use the process in normal channels. High-risk users need legal review, candidate approval, and clear disclosure.

Which AI Content Tools Are Legally Safe for Indian Election Campaigns?

Indian election law does not ban campaign teams from using AI tools. The legal risk depends on how you use the tool, what the final content claims, whether it imitates a real person, and whether voters can mistake synthetic content for real speech, real footage, or real evidence.

You can use AI for routine campaign production. You should not use AI to mislead voters, impersonate leaders, create fake events, clone voices without disclosure, or publish manipulated political content as real content.

The safest rule is:

“Use AI to improve campaign production. Do not use AI to fake political reality.”

The Election Commission of India warned political parties in May 2024 against AI-based tools to create deepfakes that distort information or spread misinformation. The ECI also directed parties to remove fake content within three hours of its coming to their notice, warn the responsible person within the party, report unlawful information and fake accounts to platforms, and escalate unresolved issues to the Grievance Appellate Committee under Rule 3A of the IT Rules.

Legal Review Note

This guide explains practical risk levels for campaign teams. It is not a substitute for legal advice. Before you generate AI-generated voiceovers, leader avatars, synthetic videos, parody clips, attack ads, or digitally recreated Scea NES scenes, get a review from a lawyer who understands Indian election law, IT law, media law, and platform policy.

Use legal review when the content includes:

• A real leader’s face

• A real leader’s voice

• An opponent’s image or speech

• A fake scene or reconstructed event

• A claim about crime, corruption, caste, religion, health, national security, or the voting process

• A parody that uses realistic impersonation

• A campaign ad that uses synthetic content without labeling

When your team is unsure, apply this test:

“Can an ordinary voter mistake this content for real footage, real audio, a real endorsement, or a real event?”

If the answer is yes, label it clearly, change the format, or do not publish it.

The Main Legal Principle

AI tools are legally safer when they help with editing, formatting, translation, accessibility, planning, and design support. They become risky when they create deception.

The ECI’s January 2025 advisory advised political parties, candidates, and campaign representatives to label syntheAI-generated campaign content. It referred to labels such as “AI Generated,” “Digitally Enhanced,“ and “Synthetic Content.”

This means your campaign should not hide synthetic media. If AI creates or significantly alters an image, audio, video, or promotional asset, add a clear label that the voter can see or hear.

Safe AI Tools for Text Drafting and Editing

Text tools are usually low risk when your team checks facts and approves the final message. You can use AI to draft speeches, captions, press notes, issue summaries, debate points, WhatsApp messages, FAQs, and social media posts.

Safe uses include:

• Grammar correction

• Translation support

• Caption drafting

• Speech outline creation

• Manifesto summary drafting

• Short video script drafting

• Email and press note drafting

• Social media post variations

• Volunteer training material

• Public issue explanation in simple language

These tools become risky when they create false claims, fake quotes, defamatory statements, religious information, or incorrect voting information.

Your team should check AI-generated text for:

• Factual accuracy

• Legal risk

• Model Code of Conduct risk

• Defamation risk

• Communal or caste sensitivity

• Gender sensitivity

• Voting process accuracy

• Source quality

Do not publish AI-generated claims without human review. AI tools can invent facts, dates, quotes, survey results, names, and legal references.

Safe AI Tools for Translation and Subtitling

Translation and subtitling tools are useful in Indian elections because campaigns work across many languages and dialects. These tools are generally safe when they preserve the meaning of the original speech.

Safe uses include:

• Translating official speeches

• Creating subtitles for videos

• Translating press releases

• Turning long speeches into short summaries

• Creating regional language captions

• Preparing accessibility captions

The risk increases when AI changes tone, adds claims, removes qualifiers, or mistranslates sensitive references. This matters in political content because one wrong word can change the meaning of a speech.

Use this workflow:

• Keep the original speech file

• Use AI for first draft translation

• Ask a native speaker to review the translation

• Check names, places, caste terms, religious terms, and policy claims

• Keep the final approved version in your archive

Do not AI-translate speeches in a leader’s cloned voice without consent, disclosure, and legal review.

Safe AI Tools for Image Editing and Design

AI image tools are safer when they create posters, backgrounds, layouts, icons, thumbnails, visual explainers, and non-realistic illustrations. They are risky when they create fake evidence.

Safe uses include:

• Poster layouts

• Background cleanup

• Image resizing

• Thumbnail variations

• Infographic design

• Icon creation

• Color correction

• Social media format adaptation

• Non-realistic illustrations

• Accessibility graphics

High-risk uses include:

• Fake rally crowds

• Fake images of leaders meeting someone

• Fake images of violence or public disorder

• Fake images of EVM tampering

• Fake newspaper screenshots

• Fake police notices or court documents

• Fake images showing an opponent doing something they did not do

Use this rule:

“If the image claims to show a real political event, it must be real.”

If AI creates or significantly alters the image, label it. The ECI’s 2025 advisory supports labeling AI-generated and synthetic campaign content.

Safe AI Tools for Video Editing

Video tools are safer when they support editing, trimming, noise cleaning, subtitle creation, and form conversion. They become risky when they change what a leader appears to say or do.

Safe uses include:

• Cutting long speeches into short clips

• Adding captions

• Removing background noise

• Adjusting brightness

• Resizing videos for platforms

• Adding title cards

• Creating explainers with stock visuals

• Creating animated policy explainers

• Adding disclaimers and labels

Risky uses include:

• Changing lip movement

• Creating fake speeches

• Showing fake events

• Reconstructing real people in fake scenes

• Creating synthetic attack ads

• Editing an opponent’s speech to change the meaning

• PrAI-generated scenes as documentary footage

If your campaign edits a speech clip, keep the full original speech link ready. This helps if someone accuses the campaign of manipulation.

Safe AI Tools for Voiceovers

AI voice tools require extra caution. Generic AI narration is safer than cloning a real leader’s voice. A cloned voice can mislead voters, especially on WhatsApp, robocalls, short videos, and local audio messages.

Lower-risk uses include:

• Generic narration that does not imitate a real person

• Accessibility voiceovers

• Explainer video narration

• Voice cleanup for a real recText-to-speechto speech for non-sensitive campaign material

Higher-risk uses include:

• Cloning a candidate’s voice

• Cloning an opponent’s voice

• Making a leader appear to say new words

• Using a celebrity or journalist voice style

• Sending AI voice messages through WhatsApp or IVR without disclosure

• Using synthetic voice in attack content

The safer rule is:

“If the voice sounds like a real person, disclose that it is synthetic unless it is an authorised real recording.”

You should also keep written consent from the person whose voice or likeness appears in the campaign asset. Without consent and disclosure, voice cloning creates legal, reputational, and platform risk.

Safe AI Tools for Chatbots and Voter FAQs

AI chatbots can help campaigns answer common questions about candidate background, manifesto points, schemes, events, and volunteer coordination. They become risky when they give wrong voting instructions, make false promises, collect sensitive data without notice, or generate unapproved political claims.

Safe chatbot uses include:

• Candidate biography FAQs

• Manifesto explanations

• Rally schedule information

• Volunteer onboarding

• Booth-level task guidance

• Public policy explainers

• Links to official campaign material

Risky chatbot uses include:

• Giving wrong polling dates or booth information

• Asking for sensitive personal data without consent

• Making religion-based voter claims

• Creating unapproved attack messages

• Generating fake claims about opponents

• Pretending to be a real candidate without disclosure

Your chatbot should clearly say it is an automated assistant. It should link to official sources of information on voting preparation. It should also block or escalate sensitive topics.

Safe AI Tools for Data Analysis and Social Listening

AI tools for monitoring public sentiment, summarizing comments, detecting viral narratives, and grouping voter concerns are generally safer than synthetic media tools. They work on analysis rather than voter deception.

Safe uses include:

• Topic clustering

• Sentiment summaries

• Comment analysis

• Media monitoring

• Rapid response tracking

• Complaint categorization

• Volunteer feedback summaries

• Constituency issue mapping

These tools still need privacy and data protection checks. Do not scrape private groups illegally. Do not use personal data in a way that violates platform terms or applicable law. Do not create sensitive voter profiles based on religion, caste, health, or private attributes.

Tools That Need Clear Disclosure

Some AI tools are inherently unsafe but require clear and visible disclosure, which you might be able to use to mislead the output into appearing as real content.

Use disclosurAI-generated leader AI-enhanced videos

• Synthetic voice AI-generated candidate avatars

• Digitally recreated historical AI-generated crowd visuaAI-generatedrated testimonials

• AI translated speeches with synthetic audio and AI-generated promotional video images used in political ads

Use labels such as:

• “AI Gene. rated..”

• “Digitally Enhanced.”

• “Synthetic Co..ntent.”

• “AI Assisted. Visual”

• “Synthetic Voiceover.”

Keep the label visible. Do not hide it in small text.

Tools You Should Avoid in Election Campaigns

Suppose significant changes create high legal and financial risk. Avoid them even if the tool allows the feature.

Do not use AI tools to create:

• Fake speeches by real leaders

• Fake videos of opponents

• Cloned voices without disclosure

• Fake endorsements

• Fake rally crowds shown as real

• Fake news bulletins

• Fake ECI notices

• Fake court orders

• Fake police documents

• Fake screenshots of media reports

• Fake communal incidents

• Fake castreligion-based statements

• Fake voting instructions

• Fake withdrawal announcements

• Fake alliance announcements

• Fake sting operation videos

• Fake evidence of bribery, crime, or violence

These formats mislead voters and expose the campaign to complaints, platform removals, legal action, and public backlash.

Platform Policy Risk

Your campaign must follow Indian law and platform rules. Meta, YouTube, X, and other platforms have policies on manipulated media, political ads, misinformation, impersonation, hate speech, and synthetic content.

Do not assume platform approval means legal safety. During the 2024 Indian election, The Guardian reported that Meta approved several AI-manipulated political ads submitted in a test by researchers, including ads containing hate speech and misinformation. The report said 14 out of 22 submitted ads were approved.

That example shows why campaigns need their own compliance review. A platform review failure does not protect your campaign from legal or political consequences.

What Your Campaign Approval Process Should Look Like

Your campaign should create a simple AI content approval workflow. Do not let local teams publish synthetic content without review.

Your workflow should include:

• Identify whether AI was used

• Identify whether the content shows a real person

• Check whether the content uses a real voice or a cloned voice

• Check whether the content claims to show a real event

• Check whether the content attacks an opponent

• Check whether the content mentions caste, religion, gender, crime, corruption, security, or the voting process

• Check whether a label is needed

• Confirm consent for face, voice, and likeness use

• Save source files and prompts

• Save final approval record

• Publish only from approved handles

Use a higher approval level for synthetic video, cloned voice, leader avatars, parody, attack ads, and anything involving opponents.

Green AI Use Cases

These AI uses are safer when humans review the final content:

• Grammar correction

• Caption writing

• Subtitle generation

• Translation drafts

• Speech outlines

• Press note drafts

• Poster layout support

• Image resizing

• Background cleanup

• Audio noise reduction

• Video trimming

• Social media format adaptation

• Accessibility captions

• Comment summarization

• Issue monitoring

• Volunteer training drafts

These uses help your team work faster without misleading voters.

Amber AI Use Cases

These uses need labels, consent, and legal review:

• Synthetic leader voiceovers

• Candidate avataAI-generated campaign videos

• AI translated speech audio

• Digitally recreated old, even AI-generated, supporter testimonials

• Parody using real lead AI-enhanced rally visuals, AI-generated images used in paid ads

Use these only when your campaign can explain how the content was made, why it is not misleading, and where the disclosure appears.

Red AI Use Cases

These uses should not be part of an election campaign:

• Fake speeches

• Opponent voice cloning

• Fake scandals

• Fake bribery videos

• Fake communal clips

• Fake caste statements

• Fake endorsements

• Fake withdrawal notices

• Fake voting instructions

• Fake ECI documents

• Fake court or police records

• Fake news screenshots

• Fake rally footage presented as real

• Deepfake attack ads

These uses do not support legitimate campaigning. They create deception.

Internal Records You Should Keep

Every campaign should keep an AI content register. This helps with legal review, platform complaints, media questions, and internal accountability.

Track these details:

• Tool used

• Content type

• Prompt or brief

• Source material

• AI output version

• Human editor name

• Legal reviewer name, if needed

• Consent record

• Disclosure label used

• Date of publication

• Platform posted on

• Approval person

• Final asset link

This record protects your campaign when someone questions whether the content was real, synthetic, edited, or authorized.

Do AI-generated political leader voices need disclosure in Indian Campaign Ads?

Yes. If your campaign uses an AI-generated voice that sounds like a real political leader, you should disclose it clearly in the ad. This applies when voters can mistake the voice for a real recording of the leader.

Indian election rules do not treat every AI voiceover as illegal. The risk depends on what the audio does—a generic narration voice carries a lower risk. A cloned voice of a real leader carries a high risk because it can mislead voters, create endorsements, or make a person appear to say words they never said.

Use this rule:

“If the voice sounds like a real political leader, tell voters that it is synthetic.”

The Election Commission of India warned political parties in May 2024 against AI-based tools to create deepfakes that distort information or spread misinformation. The ECI also directed parties to avoid publishing or circulating deepfake audio and video, false information, misleading content, and other harmful campaign material.

Why Disclosure Matters

AI voice content creates a direct trust problem. Voters often treat a voice message as personal and authentic, especially when it comes through WhatsApp, IVR calls, short videos, or local campaign groups. If the voice sounds like a known leader, many voters believe the leader actually recorded it.

That is where the legal and ethical risk starts. Your campaign should not let voters guess whether the audio is real or synthetic.

Disclosure protects your campaign because it shows:

• The campaign did not hide synthetic content

• The campaign did not impersonate a real person

• The campaign did not present AI audio as real speech

• The campaign gave voters enough context

• The campaign kept a record of responsible use

The ECI’s later advisory on synthetic campaign content moved toward labeling. It advised political parties, candidates, leaders, and campaigners to avoid using AI-generated or significantly altered content, and to use terms such as “AI-generated,” “Digitally Enhanced,“ or “Synthetic Content.”

When an AI Voice Needs Disclosure

You should disclose AI voice use when the ad includes a synthetic voice that sounds like:

• The candidate

• A party leader

• A star campaigner

• A Prime Minister, Chief Minister, MP, MLA, or local leader

• An opponent

• A celebrity endorser

• A journalist

• A community leader

• A religious figure

• A government officer

• A voter giving a testimonial

If the audio creates the impression that a real person personally recorded the message, disclose it.

Use a clear label such as:

• “AI Generated ..Voice..”

• “Synthetic Vo.iceover”

• “Voice recreated usi..ng AI..”

• “This audio uses an AI generate.d voice..”

• “This is not an original voice reco.rding”

Do not hide the disclosure in tiny text or at the end of a long caption. Put it where voters notice it.

When a Generic AI Voice Is Lower Risk

A generic AI narration voice is lower risk when it does not imitate a real person. For example, your campaign can use a neutral AI voice to explain a manifesto point, summarize a scheme, or introduce a video.

Lower risk examples include:

• A neutral narration voice in a policy explainer

• An accessibility voiceover for visually impaired voters

• A regional language narration that does not copy a leader’s voice

• A campaign video voiceover by a clearly fictional narratext-to-speech version of a public document

Even in these cases, disclosure remains safer when the ad is political, paid, or widely distributed. A short label avoids confusion.

Use this wording:

“Voiceover generated using AI. Content approved by the campaign.”

This keeps the message clear without making the ad look defensive.

When AI Voice Use Becomes High Risk

AI voice use becomes high risk when it imitates a real person or alters voters’ beliefs about a leader’s words, support, or conduct.

Avoid these uses:

• Cloning an opponent’s voice

• Making a leader appear to support your party when they did not

• Creating a fake apology, confession, or scandal audio

• Creating fake communacaste-based statements

• Creating fake voting instructions

• Creating fake withdrawal or alliance announcements

• Sending AI voice notes without disclosure

• Using a deceased leader’s voice without context and approval

• Using a celebrity voice without consent

• Using a journalist’s voice to create news-style content

These formats can trigger election complaints, platform removals, defamation concerns, impersonation concerns, and public backlash.

Use this internal rule:

“Never use AI voice to make a real person say something they did not say.”

What the ECI Position Means for AI Voice Ads

The ECI’s May 2024 direction focused on responsible and ethical use of social media by political parties. It warned against deepfake audio and video, misinformation, false or misleading information, derogatory content against women, the use of children in campaigns, and depictions of violence or harm to animals.

For AI voice ads, this means your campaign should avoid synthetic audio that:

• Distorts facts

• Misleads voters

• Imitates a real person without clarity

• Creates fake speech

• Damages an opponent through fabricated audio

• Spreads wrong voting information

• Creates public order risk

• Targets caste, religion, gender, or community identity

The ECI also directed parties to remove fake content within three hours of its coming to their notice, when such content falls within their control. It asked parties to warn responsible persons, report fake accounts and unlawful information to platforms, and escalate unresolved issues to the Grievance Appellate Committee under Rule 3A of the IT Rules.

Consent Is Separate From Disclosure

Disclosure alone is not enough when the AI voice imitates a real person. You also need consent.

If your campaign uses the candidaAI-generated voice, get written approval from the candidaAI-generated campaign office. If you use a party leader’s voice, get written approval from that leader or the party authority overseeing campaign communications.

For any AI voice, keep records of:

• Script approved

• Person whose voice was used

• Consent document

• Tool used

• Date of generation

• Final audio file

• Disclosure text

• Platform where it was published

• Person who approved publication

This record helps if the latter subject faces a competitor.

AI Voice of Your Own Candidate

Using your own candidaAI-generated voice is not automatically unsafe. It becomes safer when you have consent, script approval, and disclosure.

Safe practice:

• Get written approval from the candidate

• Use only approved scripts

• Add a clear disclosure

• Do not create emotional or false personal claims

• Do not use the voice for sensitive claims without legal review

• Keep the original script and audio file

• Keep the approval record

Example disclosure:

“This ad uses an AI generated voice approved by the candidate.”

This tells voters that the message belongs to the campaign, while also making clear that the audio is not a live or original recording.

AI Voice of an Opponent

Do not use an AI-generated voice in campaign ads. This is one of the highest-risk uses of AI in elections.

Even if your team calls it satire, voters can mistake it for real audio. A realistic opponent voice can pose risks of false attribution, defamation, impersonation, and election misinformation.

Avoid:

• Fake opponent speeches

• Fake opponent phone calls

• Fake apology audio

• Fake scandal audio

• Fake communal statement audio

• Fake alliance announcement audio

• Fake candidate withdrawal audio

If you want to criticize an opponent, use your own narrator, verified quotes, public records, or original video clips with proper context.

AI Voice in Satire and Parody

Satire and parody need careful handling. Indian election rules do not have a clear rule that protects every AI parody. If the voice sounds real, voters can misunderstand it.

Safer parody use of cartoon-style voices

• Clearly fictional characters

• Obvhumorumour format

• Visible and audible disclaimer

• No realistic imitation of a real leader

• No fake news format

• No false factual claim

Risky parody uses:

• Realistic cloned leader voice

• Fake breaking news style audio

• Fake phone call format

• Fake confession or scandal format

• Fake communal speech

• Fake voting instruction

Use this test:

“Will a normal voter understand this as parody within the first few seconds?”

If not, do not publish it.

AI Voice in WhatsApp, IVR, and Robocalls

AI voice needs more care in WhatsApp, IVR, and phone campaigns because voters hear the message without visual context. A short audio clip can spread without captions, disclaimers, or source details.

For WhatsApp voice notes, include a written caption:

“AI generated voiceover. Official campaign message.”

For IVR or robocalls, add an audio disclosure at the start:

“This is an AI-generated voice message approved by the campaign.”

Do not place the disclosure only at the end. Many users stop listening early.

For short videos, add a bobot on-screen disclosure

• Audio disclosure or caption disclosure

What MeitY Rules and Advisories Mean

MeitY’s framework matters because AI voice content spreads through online intermediaries and platforms. The IT Rules, 2021, regulate intermediaries’ duties regarding online content and grievance handling. MeitY has also issued advisories to intermediaries on platform AI-related risks, misinformation, and deepfakes.

In October 2025, MeitY proposed amendments to address deepfakes and synthetic content through clear labeling, traceability, and accountability measures. The explanatory note said MeitY had earlier issued multiple advisories to curb deepfake content and related harms.

For campaign teams, the direction is clear: synthetic political media is moving toward stricter disclosure and traceability. Your campaign should adopt a disclosure approach rather than wait for enforcement issues.

Platform Policy Risk

Even if Indian law does not give a single AI voice rule for every case, platforms have their own rules on manipulated media, political ads, synthetic content, impersonation, and misinformation.

Your campaign should assume that a political ad using a singing AI voice can face removal or rejection when it:

• Imitates a real person without disclosure

• Misleads users about who is speaking

• Claims a fake endorsement

• Uses manipulated media to attack an opponent

• Gives wrong voting information

• Creates public safety or communal risk

• Hides synthetic content from viewers

Do not treat platform approval as legal clearance. During the 2024 Indian election, published reporting found gaps in the platform’s review of AI-manipulated political ads, including Meta’s ad approvals in a test involving AI-manipulated political content.

Recommended Disclosure Wording

Use simple wording. Do not use vague labels like “AI assisted“ when the voice itself is synthetic.

Fcandidate-approved ad:

“This ad uses an AI generated voice approved by the candidate.”

For a pleader’s message:

“Synthetic voiceover used with campaign approval.”

For a generic narration:

“Voiceover generated using AI.”

For a regional language translation:

“AI generated translated voiceover. Original message approved by the campaign.”

For parody:

“Satirical AI generated voice. Not a real recording.”

For recreated historical content:

“Synthetic voice used for creative recreation. Not an original recording.”

The disclosure should appear in the caption, on screen, or in the audio, depending on the format. For audio-only formats, say it at the beginning.

How ECI and MeitY Guidelines Regulate Deepfakes in Indian Elections

India regulates election deepfakes through two main routes. The Election Commission of India controls election conduct, campaign responsibility, and political party duties. Moverseestrols digital platform duties, intermediary compliance handling of nce, unlawful codling, and online harms under the IT framework.

There is no single Indian election only for deepfakes. You need to read ECI directions, MeitY advisories, the IT Act, the IT Rules, the Representation of the People Acts, the Model Code of Conduct, criminal law, and platform policies together.

Use this working rule:

“ECI regulates how political actors use or respond to deepfakes during elections. MeitY regulates how platforms and intermediaries handle deepfakes online.”

For your campaign, this means two things. First, do not create or share deceptive AI content. Second, if someone targets your candidate with a deepfake, preserve proof, report it to platforms, escalate it through election channels, and issue a clear public correction.

What the ECI Regulates

The Election Commission of India focuses on election fairness, campaign conduct, and political party responsibility. During the 2024 Lok Sabha elections, the ECI warned political parties against AI-based tools to create deepfakes that distort information or spread misinformation. It also reminded parties that the existing legal framework already covers misinformation, impersonation, synthetic content, and campaign misconduct. This includes the Information Technology Act, 2000, the IT Rules, 2021, the Indian Penal Code, the Representation of the People Acts, and the Model Code of Conduct.

The ECI directed political parties to avoid:

• Deepfake audio and video

• False or misleading information

• Misinformation that distorts facts

• Derogatory content against women

• Use of children in political campaigning

• Depictions of violence or harm to animals

• Unlawful content shared through fake accounts

• Social media content that violates the Model Code of Conduct

For your campaign, this means AI-generated content cannot serve as a cover for fake speech, fake videos, fake endorsements, fake voting information, or manipulated attack material.

The EThree-Hour Hour Removal Rule

The ECI gave political parties a direct operational duty during the 2024 elections. If fake content comes to a party’s notice and the party controls that content or its distribution channel, the party must remove it within three hours. The ECI also asked parties to warn the responsible person within their party, report unlawful information and fake accounts to the platforms, and escalate unresolved issues to the Grievance Appellate Committee under Rule 3A of the IT Rules.

This applies to your campaign when fake or manipulated content appears through:

• Official party pages

• Candidate pages

• Local party units

• Party worker accounts

• Volunteer groups

• Campaign managed WhatsApp or Telegram groups

• Linked influencer page District-level social media handles

• IT cell distribution networks

You should not wait for public criticism before acting. Once your team finds the fake content, remove it, warn the person who shared it, preserve proof, and record your action.

What the ECI Says: A Labeling AI Content

The ECI moved from takedown responsibility to disclosure responsibility after the 2024 election cycle. In its 16 January 2025 advisory, the ECI advised political parties to label synthetic AI-generated content used for election campaigning. The advisory covered images, video, audio, and other campaign material generated or significantly altered by AI. It referred to labels such as “AI Generated,” “Digitally Enhanced,“ and “Synthetic Content.”

For campaign teams, this creates a clear publishing standard. If AI creates or significantly changes a campaign asset, label it. Do not hide the label. Do not place it where voters cannot see it.

Use disclosure of AI-generated leader images

• Synthetic voice AI-generated campaign videos

• Digitally altered leader visuals

• AI translated speeches using synthetic audio and AI-generated rally visuals using real political figures

• Synthetic public service style aAI-generated testimonials or endorsements

The safest label is direct:

“AI Generated Content”

For a synthetic voice, use:

“AI Generated Voice”

For an edited video, use:

“Digitally Enhanced Video”

What MeitY Regulates

MeitY regulates the digital and platform side of deepfakes. It is set out in the Information Technology Act, the IT Rules, and advisories to intermediaries. Intermediaries include social media platforms, video platforms, hosting services, and other online services that carry user content.

MeitY issued advisories to intermediaries, urging them to comply with the IT Rules and address deepfakes, misinformation, false content, and impersonation. A December 2023 PIB release on MeitY’s advisory said platforms should identify and remove misinformation, false or misleading content, and material impersonating others, including deepfakes.

For your campaign, MeitY’s framework matters when you report content to platforms. You need to give platforms clear reasons why the content violates rules. A complaint should state whether the content involves:

• Impersonation

• Deepfake audio or video

• False or misleading information

• Defamation

• Election misinformation

• Fake account activity

• Hate speech

• Public order risk

• Synthetic media without disclosure

How the IT Rules Affect Platforms

The IT Rules, 2021 set due diligence duties for intermediaries. These include coverage grievance handling responses, including unlawful responses to user complaints, and platform accountability. The rules are not only for elections; they become highly relevant during elections because deepfakes spread on online platforms.

The ECI itself linked election deepfake concerns to the IT Act and IT Rules in its May 2024 direction. That matters because campaign teams can use election authority channels and platform complaint channels to respond to deepfake attacks.

Your complaint to a platform should include:

• Link to the fake content

• Screenshot

• Screen recording

• Account handle

• Date and time of discovery

• Why the content is fake or manipulated

• Candidate or party affected

• Original video or audio, if available

• Election harm caused

• Request for removal or action

Do not send a vague complaint such as “This is fake.“ Give proof.

MeitY’s Direction on DeepLabeling and Traceability

MeitY’s policy direction has moved toward cleantech labeling, traceability, and platform accountability for synthetic content. In October 2025, MeitY released an explanatory note for proposed amendments to the IT Rules. The note said MeitY had earlier issued multiple advisories to intermediaries, including social media intermediaries and significant social media intermediaries, to curb deepfake content and related harms. It also said the proposed amendments aimed to create a clearer legal basis for labeling, traceability, and accountability for synthetically generated information.

Reuters reported that the proposed rules required platforms to label AI-generated visual media with a marker covering at least 10 percent of the image, and to label audio clips with an identifier during the first 10 percent of playback. It also reported that users would need to declare whether uploaded content is AI-generated, and companies would need technical systems for verification.

For campaign teams, the policy direction is clear. You should not wait for enforcement to become stricter. Slabeling synthetic content now. Keep records. Use AI disclosure in voice, video, image, and campaign ad formats.

How ECI and MeitY Work Together

ECI and MeitY regulate different parts of the same problem.

The ECI focuses on political actors and election conduct. It tells parties what they should not publish, how fast they must remove fake content under their control, and how they should report or escalate unlawful content during elections.

MeitY focuses on platforms and intermediaries. It sets online due diligence expectations and rules for responding to unlawful responses. It also pushes platforms to labeling and traceability for synthetic media.

For your campaign, both routes matter.

If your own team shares a deepfake, the ECI route creates campaign responsibility. If an anonymous account spreads a deepfake about your candidate, the MeitY route helps you report it to platforms and escalate the complaint. If the content affects voters during an election, use both routes together.

What Counts as a Deepfake in Campaign Practice

A deepfake is just a face-swap video. In election work, treat any synthetic or manipulated content as high risk when it makes a real person appear to say, do, support, or endorse something they did not say, do, support, or endorse.

Common election deepfake formats include:

• AI is the one voice of a candidate

• Fake video of a leader making a speech

• Eda edited a lip that changes the meaning of a real speech

• Fake endorsement by a public figure, an AI-generated image of a meeting that never happened

• Fake withdrawal announcement

• Fake alliance announcement

• Fake communal statement

• Fake voting instruction

• Fake scandal audio

• news-style video using synthetic visuals

Use this test:

“Does this content create a false belief about a real person, real statement, real event, or real endorsement?”

If yes, treat it as a deepfake or manipulated media risk.

What Campaigns Must Not Do

Your campaign should not publish or circulate AI content that misleads voters or impersonates real people. This rule applies even if the content performs well online.

Avoid:

• AI cloned opponent voices

• Fake candidate speeches

• Fake leader endorsements

• Fake news graphics

• Fake ECI notices

• Fake court or police documents

• Fake rally footage presented as real

• Fake crowd visuals

• Fake communal statements

• Fake voting process instructions

• Edited clips that distort an opponent’s statement, AI-generated attack ads that look like real evidence

If your campaign cannot prove that a factual claim is true, do not publish it as fact.

What Campaigns Can Do Safely

AI tools are not banned. You can use  AI routine production as long as the content does not deceive voters.

Lower risk uses include:

• Caption drafting

• Grammar correction

• Translation support

• Subtitle generation

• Video resizing

• Poster layout

• Background cleanup

• Audio cleanup

• Accessibility captions

• Speech outline support

• Social media format adaptation

• Policy explainers

• Data visualization

• Monitoring and issue summaries

These uses help production. They do not fake political reality.

Higher risk uses need consent, labels, and legal review

• Candidate avatAI-generated campaign video

• AI translated speech in synthetic voice

• Parody using real leader likeness

• Digitally recreated historical scenes, AI-generated testimonies, and AI-generated attack visuals

Use this rule:

“If voters can mistake it for real, label it or do not publish it.”

How to Respond When a Deepfake Targets Your Candidate

When a deepfake targets your candidate, you need a fast response that uses both ECI and MeitY routes.

First, save proof. Capture the URL, screenshots, screen recording, account name, handle, caption, date, time, comments, and engagement count.

Second, verify the content. Check whether the iAI-generated, edited, old footage, mistranslated, or taken out of context.

Third, report it to the platform. Use categories such as impersonation, manipulated media, election misinformation, defamation, hate speech, or fake account activity.

Fourth, escalate to election authorities if the content affects voters, impersonates a candidate, or violates campaign rules.

Fifth, issue a public clarification from verified handles. Do not repost the full deepfake in high quality. Use a blurred screenshot or warning frame.

Use this wording:

“This video is fake and manipulated. The candidate did not make this statement. We have reported it to the platform and election authorities. Please do not forward it.”

Sixth, brief journalists, district coordinators, party workers, WhatsApp admins, and local influencers. Tell them what is fake, what is true, and which official link to share.

How to File a Strong Platform Complaint

A strong complaint helps platforms act faster. Your complaint should be short, factual, and complete.

Include:

• “This is a deepfake or manipulated media comp..laint..”

• Candidate name

• Party name

• Constituency

• Link to the fake content

• Account handle

• Screenshot and screen recording

• Why the content is false

• Real source material, if available

• Election harm caused

• Request for urgent removal

• Contact details for follow-up

Do not add emotional language. Focus on proof.

How to File an Election Complaint

When a deepfake affects election conduct, file a complaint with the relevant election authority. Your legal team should attach a short evidence note.

Include:

• Date and time of discovery

• Platform where the content appeared

• Link and screenshots

• Candidate affected

• Nature of manipulation

• How it misleads voters

• Whether it uses cloned voice, fake video, or fake claims

• Whether it targets caste, religion, women, children, public order, or the voting process

• Platform complaint details

• Requested action

Connect the complaint to the ECI’s May 2024 direction where relevant, especially if the content involves deepfake audio or video, false information, fake accounts, or party-controlled or party-controlled circulation.

Deepfake, Satire, and Political Parody Under Indian Election Rules

Indian election rules do not provide a single, clear, neat definition that distinguishes rates every deepfake from satire or political parody. You need to judge the content by its effect on voters, its realism, its factual claim, and its risk to election fairness.

Deepfakes pose a legal and campaign risk when they lead voters to believe a real person said, did, supported, or endorsed something that never happened. Satire and parody can carry lower risk when voters can easily understand that the content is fictional, humorous, or critical commentary—the parody starts when satire or parody appears to be like real evidence.

Use this test:

“Can an ordinary voter mistake this content for a real statement, real event, real endorsement, or real recording?”

If the answer is yes, treat it as high risk.

The Election Commission of India warned political parties in May 2024 against AI-based tools to create deepfakes that distort information or spread misinformation. The ECI also directed parties to remove fake content within three hours of its coming to their notice, warn the responsible person within the party, report unlawful information and fake accounts to platforms, and escalate unresolved issues to the Grievance Appellate Committee under Rule 3A of the IT Rules.

What Counts as a Deepfake

A deepfake is synthetic or manipulated content that falsely presents a real person’s face, voice, body, words, or actions. In election campaigns, this includes AI-generated face-swapped video, cloned voices, manipulated images, fake speeches, and synthetic scenes that appear real.

Deepfake content usually creates a false belief about:

• What a candidate said

• What a leader did

• Who endorsed whom

• Whether a candidate withdrew

• Whether an alliance changed

• Whether a crime, scandal, or public incident happened

• Whether a community, leader, or voter group supports a party

• Whether an official notice, news report, or legal document exists

A campaign deepfake becomes especially risky when it targets behavior. For example, a fake audio message saying a candidate withdrew, a cloned voice asking voters to support another party, or a synthetic video showing a leader making a communal statement can mislead voters before corrections reach them.

Common Deepfake Examples in Elections

Treat these formats as high risk:

• AI is the one voice of a candidate

• Fake video of a leader giving a speech

• Edited speech that changes the original meaning

• Fake endorsement by a celebrity or community lead, AI-generated rally crowd shown as real

• Fake image of a political meeting that never happened

• Fake withdrawal announcement

• Fake alliance announcement

• Fake bribery or confession video

• Fake communacaste-based statement

• Fake voting instruction

• Fake ECI, court, police, or media document

Your campaign should not publish, amplify, or defend this content. If someone shares it through your campaign network, remove it and preserve proof.

What Counts as Satire

Satire criticizes politics, leaders, parties, policies, or public behavior through exaggeration, irony, or ridicule. Satire can be part of political speech when viewers understand that it is not real footage or real evidence.

Safer satire usually has these traits:

• It looks clearly fictional

• It uses cartoon, illustration, or exaggerated style

• It does not imitate a real person’s voice realistically

• It does not present fake facts as true

• It does not use fake news formatting

• It does not claim to show a real event

• It does not give false voting information

• It includes a clear label when AI is used

Example of lower risk satire:

“A cartoon style video showing fictional characters discussing campaign promises, with a clear label that it is satire.”

Example high-risk satire:

“A realistic AI video showing an opponent making a fake caste or religion based statement, later defended as satire.”

The second example creates voter deception. Calling it satire after publication does not remove the risk.

What Counts as Political Parody

Political parody imitates a style, speech pattern, slogan, campaign format, public figure, or humor, or uses it for criticism. Parody becomes safer when the average viewer understands the imitation is not real.

Safer parody uses:

• Cartoon voices

• Fictional characters

• Clehumorur format

• Visible disclaimer

• Exaggerated visuals

• No fake factual claim

• No realistic cloned leader voice

• No fake news layout

Risky parody uses:

• Realistic face swap

• Cloned political leader voice

• Fake breaking news template

• Fake phone call format

• Fake confession format

• Fake press conference format

• Fake endorsement format

• Fake voting instruction

Use this rule:

“Parody should make the joke clear before it makes the attack.”

If the viewer needs a later explanation to understand that the content was not real, your campaign has created unnecessary risk.

Where Indian Election Rules Draw the Line

Indian election rules focus less on labels, more on labels such as “satire“ or “parody“ and on harm, deception, and election conduct. If synthetic content distorts facts, impersonates a real person, misleads voters, or violates the Model Code of Conduct, it creates risk.

The ECI’s May 2024 direction brought the existing legal framework on misinformation and impersonation through deepfakes to the notice of political parties. It warned against deepfake audio and video, false or misleading information,  and misuse of AI-based tools that distort information or spread misinformation.

For campaign teams, this means you should check:

• Does the content impersonate a real person?

• Does it create a fake factual claim?

• Does it use a cloned voice?

• Does it show a fake event?

• Does it present edited content as real?

• Does it target caste, religion, women, children, violence, or public order?

• Does it mislead voters about polling, candidates, or alliances?

• Does it violate platform rules or campaign law?

If the answer is yes, do not rely on a “satdefensefence.

How Labeling Rules Affect Satire and Parody

The ECI’s 2025 advisory moved campaign practices toward labeling synthetic AI-generated content. It advised political parties, candidates, and campaigners to avoid using AI-generated or synthetic campaign material, and to use terms such as “AI-generated,” “Digitally Enhanced,“ or “Synthetic Content.”

This matters for satire and parody. If your campaign uses AI to create a parody using a synthetic voice, an edited leader image, or a digitally altered scene, add a visible label. Do not hide it.

Use labels such as:

• “Satirical AI Generated ..Video.”

• “Political P..arody.”

• “Synthetic Voice Used.”

• “Digitally Altere..d Visual.”

• “Not a Real Reco..rding.”

• “Fictional ..Scenaudio-onlyio only content, say the disclosure at the beginning. For video content, add the label on screen. For social posts, add the label in the caption and creative.

MeitY’s Role in Deepfake and Synthetic Content

MeitY regulates the platform and intermediary side of deepfakes. Its advisories and the IT Rules matter because political deepfakes spread through social media platforms, messaging apps, video platforms, and hosting services.

In December 2023, MeitY said platforms should identify and remove misinformation, false or misleading content, and material impersonating others, including deepfakes.

In October 2025, MeitY’s proposed amendments moved toward clelabeling, traceability, and accountability for synthetically generated information. The explanatory note said MeitY had earlier issued multiple advisories to intermediaries to curb deepfake content and related harms.

For your campaign, this means platform complaints should clearly explain whether the content is deepfake, impersonation, manipulated media, false information, or synthetic content without disclosure.

Safe Satire Practices for Campaign Teams

You can create satire, but you need con in placet in placerols. The safest satire does not ask voters to believe a false event.

Use these practices:

• Use cartoons or illustrations instead of realistic face swaps

• Use fictional narrator voices instead of cloned leader voices

• Add a visible satire or parody label

• Avoid fake news graphics

• Avoid fake documents, fake headlines, and fake court or police notices

• Avoid realistic scenes involving opponents

• Avoid caste, religion, gender, violence, national security, or public order claims

• Avoid fake statements about voting dates, polling booths, EVMs, alliances, or candidate withdrawal

• Keep source files, scripts, prompts, and approvals

• Get legal review of high-reach paid ads

Satire should criticize. It should not trick.

Unsafe Satire Practices

Do not use satire as a cover for misinformation. Campaign teams often take a high-risk route by creating realistic fake content and then calling it a rumor, calling it humor only after criticism starts.

Avoid:

• “Fake confession“ audio

• “Fake phone call leak“ format

• “Breaking news“ parody that looks real

• “Leaked video“ format

• Realistic cloned opponent voice, AI-generated rated leader face, saying fake lines

• Fake police, ECI, or court document jokes

• Fake endorsement jokes

• Fake communal speech jokes

• Fake bribery or criminal act visuals

These formats can mislead voters and create legal complaints.

How to Review Satire Before Publishing

Your campaign should review every satire or parody asset before publication. Use a simple approval checklist.

Ask:

• Is this clearly fictional from the first few seconds?

• Does it use a real leader’s face or voice?

• Does it create a fake quote?

• Does it look like real news or real evidence?

• Does it involve religion, caste, women, children, violence, security, or the voting process?

• Does it need an AI label?

• Do we have consent if a real voice or likeness is used?

• Can supporters misunderstand and forward it as real?

• Can opponents file a complaint with a reasonable basis?

• Did legal review clear it?

If the answer creates doubt, change the format. Use illustration, text commentary, or a fictional narrator.

How to Label Satire and Parody

Labels should be clear, visible, and early. Do not use vague wording.

Use this for a video:

“Satirical AI Generated Video. Not a Real Recording.”

Use this for an audio clip:

“This is a political parody using a synthetic voice. It is not a real recording.”

Use this for an image:

“Digitally Altered Satirical Image.”

Use this for a caption:

“Political parody. Fictional scene. AI generated visual.”

A label does not make unlawful content safe. If the content spreads false voting information, incites hostility, defames someone, or impersonates a real person, it alone does not solve the problem.

How to Treat Opponent Content

When your opponent posts satire or parody attacking your candidate, do not react emotionally. First, classify the content.

Ask:

• Is it obvious satire?

• Is it realistic deepfake content?

• Does it use your candidate’s cloned voice?

• Does it create a fake factual claim?

• Does it mislead voters?

• Does it violate ECI directions or platform rules?

• Is it spreading fast?

If the content is obvious satire, overreaction can help the opponent. If the content is realistic and misleading, save proof, report it, issue a clear correction, and escalate it.

Use this public line:

“This is a fake and manipulated clip. The candidate did not make this statement. Please do not forward it. We have reported it to the platform and election authorities.”

How to Respond to Deepfake Content Disguised as Parody

Some attackers label deepfakes as parody to reduce accountability. Your campaign should focus on voter deception, not the label used by the attacker.

Collect proof:

• Original URL

• Screenshot

• Screen recording

• Caption

• Account handle

• Date and time

• Engagement data

• Evidence of confusion among users

• Original speech or recording, if available

• Platform complaint ID

Then report it as manipulated media, impersonation, election misinformation, or deepfake content. If it affects voters during an election period, escalate it through election channels.

Practical Risk Scale for Campaign Teams

Treat campaign content as low risk when it is clearly fictitious, non-realistic, and not tied to a fake factual claim.

Treat it as medium risk when it uses real leaders, political symbols, AI visuals, or edited forms, but includes clear labeling and does not mislead voters.

Treat it as high risk when it uses realistic voices, faces, fake events, fake endorsements, fake documents, fake news presentation, or false claims about voting, caste, religion, crime, corruption, national security, or public. High-risk content needs legal review. High-risk content should not be published at all.

How Fast Meta, YouTube, and X Remove Political Deepfakes in India

There is no reliable public dataset that gives an exact average removal time for political deepfakes on Meta, YouTube, and X during India’s 2024 elections. Published reports show a clear pattern instead: platforms had policies, but enforcement was uneven, complaint-driven, and often slower than the speed at which fake content spread.

For campaign teams, the practical answer is direct:

“Do not assume platforms will remove political deepfakes fast enough to protect your candidate. Save proof, report the content, publish a correction, and escalate through election and legal channels.”

During the 2024 Lok Sabha elections, the Election Commission of India told political parties to remove deepfake audio, deepfake video, and false or misleading content within three hours of it coming to their notice when the content was under their control. The ECI also told parties to warn responsible persons, report unlawful information and fake accounts to platforms, and escalate unresolved issues to the Grievance Appellate Committee under Rule 3A of the IT Rules. This three-hour applies directly to political parties, not as a full public benchmark for Meta, YouTube, or X platform action in 2024.

The Short Answer for Campaign Teams

Meta, YouTube, and X can remove political deepfakes quickly when a complaint is clear, urgent, legally grounded, and escalated through the right channel. But in practice, many harmful posts remain live long enough to spread across WhatsApp groups, repost pages, Telegram channels, video downloads, and local media discussions.

Your campaign should work with three timelines:

• First 30 minutes: save proof, classify the fake, and alert the response team

• First 2 hours: file platform complaints, prepare legal escalation, and draft the public clarification

• First 6 hours: publish the clarification, brief journalists and local teams, and track platform action

Do not wait for the plate to be removed before you respond publicly. A fake clip can travel faster than a review queue.

Why Exact Removal Speed Is Hard to Measure

Public reports rarely show the full timeline from complaint to removal. Platforms also do not publish India-specific deepfake takedown averages for each election incident. A post can disappear because the creator deleted it, the platform removed it, a fact-check reduced its visibility, a court order forced action, or a complainant withdrew the content. These are different outcomes.

A July 2024 report by The London Story said India’s 2024 election deepfake guidance focused mainly on political candidates and parties. It also said the guidelines did not impose direct obligations on platforms and that platforms removed content only sporadically and after complaints. This matters because a campaign cannot rely on automatic platform enforcement during a live election fight.

What is required in 2024

The ECI’s May 2024 directives are clear for political parties. Parties had to avoid publishing or circulating deepfake audio and video, false information, misleading information, derogatory content against women, content using children in campaigns, and depictions of violence or harm to animals.

For fake content under party control, the ECI set a three-hour removal duty after notice. This gives campaign managers a working standard even when a platform has not acted yet.

Your campaign should treat three hours as the internal deadline for:

• Removing fake content from party-controlled pages

• Deleting deepfake posts from official handles

• Warning workers who shared fake content

• Removing fake material from campaign groups

• Reporting fake accounts to platforms

• Escalating unresolved cases through proper complaint channels

If your own network spreads a deepfake, you cannot blame the platform first. You must remove it.

What Changed for Platforms After 2024

India moved toward a faster platform after the 2024 election. Reuters reported in February 2026 that India amended its social media rules to require platforms such as Meta, YouTube, and X to remove unlawful content within three hours of notification, down from the earlier 36-hour window. The amendment was set to take effect on 20 February 2026.

This means your 2026 campaign planning should treat three hours as the expected legal response window for notified unlawful content, but you should still file a strong complaint. A weak report, lacking links, screenshots, timestamps, and legal grounds, slows action.

Meta: Policy Exists, Enforcement Gaps Remain

Meta has policies for manipulated media, political ads, AI labels, and harmful content. In April 2024, Meta said it would expand labels for AI-generated video, audio, and image content when it adopts industry-standard AI indicators or when users disclose AI generation. Meta also said it would move toward giving users more context through labels, while removing content that violates other policies.

Meta also required advertisers to disclose when AI or digital methods created or altered political and social election-related ads on Facebook and Instagram starting in 2024. Reuters reported this policy in November 2023.

But published reporting showed serious enforcement gaps during India’s 2024 election. The Guardian reported that Meta approved sevAI-manipulated political ads submitted as a test by civil society groups. The report said 14 out of 22 submitted ads were approved, including ads containing hate speech and misinformation. The researchers took the ads down before publication.

For campaign teams, this means Meta can act, but its review system does not catch every harmful AI political asset before they are approved or circulated.

How to Report Political Deepfakes on Meta

When your candidate is targeted on Facebook, Instagram, or Threads, report the post under the category that best matches it.

Use categories such as:

• Manipulated media

• Impersonation

• False information

• Hate speech

• Harassment

• Election misinformation

• Unauthorized use of image or voice

• Political ad disclosure violation

Your complaint should include:

• Post URL

• Screenshot

• Screen recording

• Page name and handle

• Date and time of discovery

• Why the content is fake

• Candidate’s official profile link

• Original speech or video link, if available

• Statement that it affects an election campaign

• Request for urgent removal or visibility reduction

Do not only say “This is fake.“ Explain what is fake and how it harms voters.

YouTube: Disclosure Rules Exist, Takedown Speed Depends on Review

YouTube requires creators to disclose when content is realistic, edited, or when viewers may think it is real. YouTube says creators must disclose content that is meaningfully altered or synthetically generated when it seems realistic. The disclosure setting appears in YouTube Studio, and labels can appear in the expanded description.

This policy helps with transparency, but disclosure is not the same as removal. If a political deepfake violates YouTube policies, your campaign still needs to report it clearly. YouTube reviews factors such as realism, disclosure, impersonation, harm, and policy violations.

For Indian campaigns, YouTube is high risk because a deepfake video can be downloaded, clipped, reposted to Shorts, shared on WhatsApp, and uploaded again through other channels. Even if YouTube removes one upload, copies can survive elsewhere.

How to Report Political Deepfakes on YouTube

When reporting a deepfake on YouTube, give YouTube a complete evidence package.

Include:

• Video URL

• Channel name and link

• Timestamp where the fake claim appears

• Screenshot of the title and description

• Screen recording of the video

• Explanation of the manipulatiinal source video, if available

• Candidate’s verified source link

• Election harm caused

• Whether the video lacks synthetic content disclosure

Use specific wording:

“This video uses manipulated or synthetic media to make the candidate appear to say something they did not say. It is being used in an election context and misleads voters.”

This is stronger than a generic complaint.

X: Fast Spread, Public Disputes, and Complaint-Based Action

X is difficult for campaign teams because posts spread fast through reposts, quote posts, screenshots, and downloads. Even if X removes or restricts one post, copies can remain active through other users.

During India’s 2024 election period, X complied with some Election Commission restrictions on political posts while publicly disagreeing with their orders. Business and Human Rights ResoCenterentre archived coverage of X complying with ECI orders while objecting to the takedown direction. This example shows that X can act after authority involvement, but platform action can include disagreement, delay, or limited visibility restrictions rather than immediate, te broad removal.

X also has a synthetic and manipulated media policy. It’s a policy stating that users may not deceptively share synthetic or manipulated media likely to cause harm, and that X may label such posts or remove them.

For campaign teams, treat X as a high-speed spread channel. Report the original post, but also monitor quote posts, screenshots, reposts, and edited versions.

How to Report Political Deepfakes on X

Your X complaint should focus on manipulation, impersonation, and election harm.

Include:

• Post link

• Account handle

• Screenshot

• Screen recording

• Time posted

• Repost and view count, if visible

• Why the media is synthetic or manipulated

• Why it harms voters

• Original video or statement, if available

• Request for removal, label, or restriction

Use this complaint line:

“This post contains synthetic or manipulated media that falsely presents the candidate as making a statement they did not make. It is likely to mislead voters during an election period.”

If the post also targets caste, religion, violence, women, or public order, state that clearly.

Platform Comparison in Plain Language

Meta has stronger ad-disclosure, stronger AI labeling, and AI labeling commitments, but reporting from India’s 2024 election showed that harmful AI-manipulated political ads still passed review in a civil society test.

YouTube has a real, irule for specific altered or synthetic content, but removal still depends on policy review, user reports, and harm assessment. A label does not automatically remove content.

X can act on manipulated media and authority orders, but the platform’s public record shows disputes over political takedown orders. Fast-repost culture also means that the takedown of one post does not stop circulation.

For all three platforms, the lesson is the same:

“Platform reporting is one part of the response. It is not the response.”

What Campaigns Should Expect in Practice

You should expect four possible outcomes after reporting a political deepfake:

• The platform removes the content

• The platform labels the content

• The platform restricts visibility

• The platform takes no immediate action

You should also expect reuploads. Attackers often change the filename, crop the video, add new music, blur the watermark, change the caption, or repost from another account.

Your monitoring team must search for:

• Same video with a different caption

• Cropped version

• Dubbed version

• Screenshot version

• Meme version

• WhatsApp forward version

• YouTube Shorts version

• Instagram Reel version

• X repost version

• Telegram upload

A single complaint does not end the incident.

Why Platform Takedown Alone Fails

Platform takedown is slower than voter-facing spread. A deepfake can move through screenshots, downloads, screen recordings, and closed groups before the original post disappears.

Your campaign needs a parallel response:

• Save proof

• Report the post

• Publish a verified denial

• Share the original evidence

• Brief journalists

• Alert local campaign teams

• Stop internal forwarding

• Ask supporters to report, not repost

• Escalate to election authorities

• Track copies and reuploads

The correction must travel through the same networks as the fake.

What a Strong Six-Hour Platform Response Looks Like

Your campaign should act in stages.

First 30 minutes

• Save all proof

• Capture URL, screenshots, and screen recording

• Identify the first visible source

• Alert legal, media, and digital teams

• Stop party workers from forwarding the fake

30 to 90 minutes

• Verify the content

• Check whether iAI-generated, rated, edited, old, or taken out of context

• Prepare platform complaints

• Prepare the public clarification

• Collect original footage or audio

90 minutes to 3 hours

• File complaints on Meta, YouTube, X, and other platforms

• Report fake accounts

• Publish clarification from verified handles

• Remove any internal copies

• Warn internal workers where needed

3 to 6 hours

• Escalate unresolved complaints

• Brief journalists and fact-checkers

• Send WhatsApp-safe clarification to district teams

• Track reuploads

• Update legal complaint file

This system gives your campaign control even when platforms move slowly.

What to Include in Every Platform Complaint

Use one standard format across platforms.

Include:

• “Political deepfake comp..laint.”

• Candidate name

• Party name

• Constituency

• Platform URL

• Account handle

• Screenshot

• Screen recording

• Date and time found

• Why the content is fake

• Why it misleads voters

• Original content link, if available

• Legal or election concern

• Request for urgent action

Add this line when relevant:

“This content affects an active Indian election campaign and falsely represents the candidate’s speech, voice, conduct, or endorsement.”

This gives the platform reviewer a clearer reason to escalate.

What Not to Do While Waiting for Removal

Do not help the deepfake spread.

Avoid these mistakes:

• Do not repost the full fake video

• Do not quote post the deepfake without blurring it

• Do not send the fake clip to every WhatsApp group

• Do not use emotional language in legal complaints

• Do not make unverified claims about who created it

• Do not wait for platform removal before publishing your clarification

• Do not let district teams create their own versions of the response

• Do not delete evidence before saving it

Use one approved public line and one approved internal instruction.

Recommended Public Clarification

Use short, direct language.

“A fake and manipulated video is being circulated in the candidate’s name. The candidate did not make this statement. We have reported the content to the platform and election authorities. Please do not forward it. Follow only verified campaign handles for official updates.”

For AI voice:

“An AI generated voice clip is being circulated as the candidate’s voice. This audio is fake. The candidate has not issued this message. Please report the clip and do not forward it.”

For fake withdrawal:

“The claim that the candidate has withdrawn is false. The candidate remains in the election. Please ignore fake posts and follow verified updates.”

How to Judge Whether Platform Action Was Fast Enough

Your campaign should measure platform response in practical terms, not only by removal time.

Track:

• Time of first detection

• Time of first complaint

• Complaint reference number

• Time of platform response

• Whether content was remolabeled or restricted

• Whether copies remained active

• Number of reuploads found

• Whether the fake reaches the media or WhatsApp groups

• Time of official clarification

• Reach of your correction

If the platform removes one post after eight hours, but the fake has already moved to 100 WhatsApp groups, the response failed from a campaign perspective.

How IT Cells Can Detect and Debunk Deepfake Videos During Election Campaigns

Deepfake videos create a direct risk for Indian political campaigns. A fake speech, cloned voice, edited rally clip, or synthetic scandal video can spread across WhatsApp groups, Instagram Reels, Facebook pages, YouTube Shorts, X posts, Telegram channels, and local media pages within minutes.

Your IT cell needs a clear system. Do not wait until a fake video goes viral. Build a detection, verification, reporting, and debunking workflow before campaign season starts.

Use this working rule:

“Detect early, save proof, verify fast, report properly, and correct the record through trusted channels.”

The Election Commission of India warned political parties in May 2024 against AI-based tools to create deepfakes that distort information or spread misinformation. The ECI also directed parties to remove fake content within three hours of its coming to their notice, warn the responsible person within the party, report unlawful information and fake accounts to platforms, and escalate unresolved issues to the Grievance Appellate Committee under Rule 3A of the IT Rules.

Why IT Cells Need a Deepfake Response System

An IT cell cannot treat deepfakes as normal negative content. A normal criticism post attacks your candidate’s image. A deepfake attacks public trust by making voters believe something false is real.

Deepfake videos can create confusion around:

• Candidate speeches

• Party alliances

• Candidate withdrawal claims

• Communacaste-based statements

• Fake corruption allegations

• Fake sting videos

• Fake voting instructions

• Fake endorsements

• Fake private conversations

• Fake apology or confession clips

Your IT cell must separate genuine criticism from manipulated media. Do not report every negative post as fake. That weakens your credibility. Focus on content that changes facts, impersonates real people, or presents synthetic content as real evidence.

Set Up a Monitoring Desk

Your IT cell staff will run a monitor throughout the entire campaign period. The desk should track public platform language channels, party worker reports, and, where possible, closed-group and closed-group alerts.

Monitor these sources:

• X posts and replies

• Facebook pages and groups

• Instagram Reels and comments

• YouTube videos and Shorts

• WhatsApp group alerts from booth teams

• Telegram channels

• Local meme pages

• District news pages

• Influencer handles

• Opposition campaign pages

• Search results for candidate names

• Regional language keywords

• Nicknames and misspellings

• Fake quote cards and edited video captions

Track keywords in English, Hindi, Telugu, Tamil, Kannada, Bengali, Marathi, Malayalam, and the main language of your constituency. Many fake videos do not use official names. They use nicknames, caste references, local slogans, or short insult phrases.

Create a Suspicious Content Intake Form

Do not ask workers to forward suspicious videos in every WhatsApp group. That spreads the fake. Give them one reporting link or one internal number.

Your intake form should ask for:

• Platform name

• Post link

• Screenshot

• Screen recording, if possible

• Date and time found

• Account name

• Caption text

• Language used

• Group or page where it appeared

• Whether voters are asking about it

• Whether local media has picked it up

• Whether the content claims to show a real speech, event, or statement

Tell workers:

“Do not forward the fake video. Send the link and screenshot only to the monitoring team.”

This reduces accidental spread.

Build a Candidate Media Archive

Deepfake verification becomes faster when your IT cell has a clean archive of real content. You need original videos, speeches, photos, audio clips, rally footage, interviews, and press conferences in one searchable folder.

Your archive should include:

• Full speeches

• Rally videos

• Press conference recordings

• Interview clips

• Candidate photos

• Voice recordings

• Public statement transcripts

• Schedule records

• Event location details

• Official social media posts

• Approved campaign ads

• Media coverage links

When a fake video appears, your team can compare it with real footage. This helps determine whether the clip was edited, taken out of context, dubface-swapped, or fully generated.

Check the First Signs of Manipulation

Your IT cell should train a small review team to spot common warning signs. These signs do not prove a deepfake on their own, but they help you decide whether to escalate the video for further review.

Check for:

• Lip movement that does not match the words

• Strange eye movement

• Blurry edges around the face

• Skin texture changes near the mouth

• Lighting mismatch between face and background

• Voice tone that changes unnaturally

• Robotic pronunciation

• Missing breath sounds

• Sudden audio cuts

• Background noise that does not match the scene

• Cropped video that hides context

• Fake subtitles that change meaning

• Low resolution around the face while the background stays sharper

• Repeated frames or unnatural head movement

• Video posted with the source

Do not publish a technical claim unless you can support it. Say “fake and manipulated” when you know the statement is false. Say “under verification” only in the first internal stage, not as your final public line.

Classify the Video Before Responding

Your team should classify every suspicious video before deciding on the response.

Edited Real Clip

A real video has been cut, rearranged, slowed down, captioned falsely, or taken out of context.

Respo: Release the full original clip with a clear explanation.

Dubbed or Voice Replaced Clip

The video is real, but someone replaced the audio or added fake subtitles.

Response: publish the original audio and mark the fake version as manipulated.

AI Face Swap

The video uses a real or synthetic body and places the candidate’s face on it.

Response: report as manipulated media and impersonation. Share proof through official handles.

AI Cloned Voice

The video or audio uses a synthetic voice that sounds like the candidate.

Respo: Clarify that the voice is fake. Report it as impersonation and synthetic media.

FAI-Generated Video

The video creates an event, speech, or scene that never happened.

Response: state that no such event took place. Share the candidate’s official schedule or original records.

Parody Presented as Real

The content malabeledellehumorumour in one place but spreads elsewhere as real.

Respo: Clarify that the video is fictional or manipulated, especially when voters are confused.

Use a Risk Scale for Response Priority

Not every fake video needs the same level of response. Your IT cell would classify.

Low Risk

• Obvious joke

• Low reach

• No realistic impersonation

• No fake factual claim

• No voter confusion

Action: monitor, archive, and report if needed.

Medium Risk

• Edited speech

• Fake caption

• Local page sharing

• Some voter confusion

• Opposition supporters amplifying it

Action: publish clarification, report platform links, and brief local teams.

High Risk

• Realistic deepfake video

• Cloned candidate voice

• Fake withdrawal claim

• Fake alliance claim

• Communacaste-based statement

• Fake scandal or bribery clip

• Viral WhatsApp spread

• Media pages sharing it

• Paid ad or coordinated amplification

Action: public denial, platform complaint, legal escalation, ECI complaint, journalist briefing, district-level counter-distribution.

Verify the Video With a Step-by-Step Process

Your IT cell should follow the same verification process every time.

Start with source tracing. Find the earliest visible post. Check whether the account is new, anonymous, politically linked, or part of a recurring pattern of misinformation.

Then check the claim. Ask what the video wants voters to believe. Does it claim a real speech, real event, real endorsement, real scandal, or real announcement?

Next, compare with your archive. Search for the original event, background, clothing, mic setup, stage, banner, and speech line. Many fake videos use real footage with altered audio.

Check the audio. Listen for unnatural pauses, repeated tone patterns, missing room sound, or a mismatch between the speaker and background noise.

Check the subtitles. Fake videos often use subtitle makers to create a false claim,m even when the audio otherwise.

Check metadata only when available. Social media platforms often strip metadata, so do not depend on it.

Ask a forensic expert when the content is high risk. Your IT cell performs first-level verification, but a serious deepfake accusation requires a more thorough review.

Preserve Evidence Before Reporting

Do not report or comment before saving proof. Platforms, attackers, or page admins can delete the video. You still need evidence for complaints, legal review, media briefing, and internal records.

Save:

• Original URL

• Screenshot of the post

• Screen recording of the full video

• Account name and handle

• Profile URL

• Caption and hashtags

• Comments showing confusion

• View count, share count, and likes

• Time and date of discovery

• Names of pages that reposted it

• WhatsApp screenshots, if available

• Telegram links, if available

• Downloaded copy for legal review

Create one folder for each incident. Use a clear file that includes the date, platform, candidate name, and content type.

Example:

“2026-05-19_X_DeepfakeVideo_CandidateName_FirstUpload”

Report the Video to Platforms

Your IT cell should file platform complaints with complete proof. A weak complaint slows removal.

Use categories such as:

• Manipulated media

• Impersonation

• Election misinformation

• False information

• Defamation

• Hate speech

• Harassment

• Fake account

• Unauthorized use of image or voice

• Synthetic media without disclosure

Your complaint should say:

“This video uses manipulated or synthetic media to falsely show the candidate making a statement they did not make. It is spreading during an election campaign and can mislead voters.”

Add links, screenshots, screen recordings, and original proof. If the same video appears on multiple platforms, report each version separately.

Published reports from the 2024 Indian election showed gaps in platform enforcement. The Guardian reported that Meta approved 14 of 22 AI-manipulated political ads submitted as part of a civil society test, including those containing hate speech and misinformation. The researchers removed the ads before publication. This shows why campaign teams should not rely only on platform review.

Escalate Through Election Channels

If the deepfake affects voters, impersonates your candidate, or violates campaign rules, escalate it to the relevant election authority.

The ECI’s May 2024 direction specifically warned parties against deepfake audio and video, false or misleading information, and AI-based content that distorts information or spreads misinformation. It also requires the immediate removal of such content within three hours after notice when the party has control.

Your election complaint should include:

• Candidate name

• Party name

• Constituency

• Platform links

• Screenshots

• Screen recording

• Account handle

• Date and time found

• Nature of manipulation

• Election harm caused

• Whether the content involves caste, religion, women, violence, public order, or the voting process

• Platform complaint reference numbers

• Request for action

Keep the complaint factual. Do not add political attacks.

Publish a Clear Debunk From Verified Handles

Do not wait for the plate to be removed before correcting the record. Your official clarification should come from verified handles and trusted campaign channels.

Use short language.

For a fake video:

“A fake and manipulated video is being circulated in the candidate’s name. The candidate did not make this statement. Please do not forward it. We have reported it to the platform and election authorities.”

For fake audio:

“An AI generated voice clip is being circulated as the candidate’s voice. This audio is fake. The candidate has not issued this message. Please report it and do not forward it.”

For an edited speech:

“The viral clip cuts and changes the meaning of the speech. Watch the full original video here. Please share only verified content.”

Do not repost the full fake video in high quality. Use a blurred screenshot, a warning label, or a short, cropped frame only when needed.

Use Original Proof to Debunk Faster

A denial works better when you show the real source. Your IT cell should publish original proof whenever possible.

Use:

• Full original speech video

• Official event recording

• Correct transcript

• Candidate schedule

• Location proof

• Media coverage from the real event

• Official statement

• Behind-the-scenes footage, if useful

• Full clip before and after the edited portion

Your debunk should answer three questions:

• What is fake?

• What is true?

• What should voters do now?

Example:

“The viral video uses fake audio on an old rally clip. The original speech is here. The candidate did not make the statement in the fake video. Do not forward the manipulated clip.”

Brief Journalists and Fact-Checkers

Do not assume journalists will see your social post. Send a short note to political reporters, local media desks, fact-checkers, and district correspondents.

Your media note should include:

• One-line summary

• Link to fake content, if safe to share privately

• Screenshots and evidence

• Link to official clarification

• Link to original speech or proof

• Platform complaint status

• Campaign spokesperson contact

• Clear quote from the candidate or campaign

Use this quote format:

“The video is fake and manipulated. The candidate did not make this statement. We have reported it to the platform and election authorities. We request voters to rely only on verified campaign channels.”

Keep it clean. Do not over-explain.

Control Internal Party Distribution

Party workers often spread the fake while trying to expose it. Stop this early.

Send this instruction to all campaign groups:

“Do not forward the fake video. Share only the official clarification link. Report the original post on the platform.”

Your internal instruction should say:

• Do not repost the fake video

• Do not add angry captions

• Do not tag media with the full fake clip

• Do not create your own debunk graphics

• Do not claim who made it unless confirmed

• Do not share unverified forensic details

• Use only approved campaign language

• Share the official clarification link

This gives your campaign one voice.

Use Local Language Debunks

Deepfakes often spread in local languages. Your debunk must match the language of the attack.

Create short versions in:

• English

• Hindi

• State language

• District dialect where needed

• WhatsApp-friendly plain text

• Image format

• Short video format

• Audio clarification format

If the fake video spreads in Telugu, a long English statement will not stop it. Use the same language voters saw in the fake.

Create a WhatsApp-Safe Debunk

WhatsApp is one of the hardest channels to control because content moves through closed groups. Your debunk should be short, direct, and easy to forward.

Example:

“The video claiming that [candidate name] said [false claim] is fake. The audio has been manipulated. Please do not forward it. Official clarification: [link].”

Add a small image with:

• “Fake Video.”

• “Do Not Forward.”

• “Official Clarification.”

• Candidate verified handle

• Link or QR code, if useful, both teams to share the clarification, not the fake clip.

Train Workers to Spot Deepfake Warning Signs

Your IT cell should train district admins and booth-level digital volunteers before the forthcoming polling season.

Teach them to watch for:

• Sudden viral video from an unknown account

inal source link

• Strange lip movement

• Odd voice tone

• Mismatched subtitles

• Short clip without context

• Fake breaking news style

• New accounts posting only attack content

• Same video shared with different captions

• Claims about withdrawal, alliance, religion, caste, scandal, or voting process

Give them one instruction:

“Report suspicious content to the IT cell. Do not forward it to voters.”

Use AI Detection Tools Carefully

AI detection tools can help, but they are not final proof. Detection tools can make errors, especially with compressed social media videos, reuploaded clips, low-quality footage, and dubbed content.

Use AI detection tools for:

• First-level triage

• Pattern checking

• Technical clues

• Prioritizing high-risk content

• Supporting forensic review

Do not publish solely because one tool said the video was AI-generated. The tool’s output includes human review, original comparisons, and content analysis.

Use this public wording when technical proof is still being reviewed:

“The video is fake and manipulated. The candidate did not make this statement.”

This focuses not only on the false claim but also on the technology.

Maintain a Deepfake Incident Log

Every incident should become a case file. This helps you find repeat attackers, report coordinated activity, brief lawyers, and improve future response.

Your incident log should track:

• Incident ID

• Date and time found

• Platform

• Account handle

• Content link

• Candidate affected

• Claim made

• Type of manipulation

• Risk level

• Screenshots saved

• Platform complaint filed

• Complaint reference number

• ECI or legal complaint filed

• Public clarification published

• Journalist note sent

• Current status

• Reuploads found

• outcome

This record also helps when fake content resurfaces days later.

Create Pre-Approved Debunk Templates

Speed matters. Your IT cell should not start from zero during a crisis.

Keep templates ready for:

• Fake video

• Fake audio

• Fake withdrawal claim

• Fake alliance claim

• Fake communal statement

• Fake caste statement

• Fake corruption allegation

• Fake endorsement

• Fake ECI notice

• Fake voting instruction

Follow ECI Labeling Standards for Your Own AI Content

Your IT cell should also control the use of AI. If your team uses AI-generated images, videos, voices, or synthetic content, clearly label it.

The ECI has advised political parties to use AI-generated campaign content with notations such as “AI Generated,” “Digitally Enhanced,“ or “Synthetic Content.”

This matters because your campaign loses credibility if it complains about deepfakes while publishing unclear synthetic content itself.

Use labels such as:

• “AI Generated Video.”

• “Synthetic Voiceover.”

• “Digitally Enhanced Imag..e.”

• “Political Parody, Not a Real Recording.”

• “AI Generated Visual.”

Do not hide labels in tiny text.

Build Internal Approval Rules for AI Content

Your IT cell should not allow local teams to create AI content without review. Set clear rules before campaign work begins.

Require approval for:

• AI-generated leader video

• AI-generated leader voice

• Candidate avatar

• Parody using real leader likeness

• Attack ads using edited video

• AI-generated images of rallies

• AI-generated testimonial content

• AI translated speech audio

• Synthetic visuals used in paid ads

Ban these formats:

• Opponent voice cloning

• Fake speech videos

• Fake endorsement clips

• Fake withdrawal announcements

• Fake voting instructions

• Fake ECI documents

• Fake police or court records

• Fake scandal videos

• Fake communal or caste-based clips

Use this internal rule:

“No synthetic political content goes live without label, approval, and saved source records.”

Coordinate With Legal and Compliance Teams

Your IT cell should not handle deepfakes alone. Legal requirements when a case involves defamation, public order, electoral, or platform policy issues.

Involve legal teams when the content includes:

• Candidate impersonation

• Cloned voice

• Fake criminal allegation

• Communal or caste-based claim

• Women targeted with derogatory content

• Children shown in campaign material

• Violence or public disorder

• Fake voting information

• Fake official documents

• High-reach platform spread

• Media amplification

Give legal teams a clean evidence folder. Do not send scattered screenshots across chat groups.

Measure Whether the Debunk Worked

A debunk is not complete when you post it. Track whether the correction reached the affected audience.

Measure:

• Views on the clarification post

• Shares by local teams

• Media corrections published

• Platform removals

• Number of reuploads found

• WhatsApp reports after clarification

• Search volume for the fake claim

• Influencers still sharing the fake

• Voter questions received by field teams

If the fake keeps spreading, publish a second clarification with stronger proof. Use video, audio, and local language formats.

What Indian Campaign Managers Must Know About AI-Generated Election Content Compliance

Indian campaign managers can use AI in election content, but they need rules, records, labels, and approval controls. AI tools are useful for writing, translation, subtitles, design, video editing, audio cleanup, monitoring, and content planning. The risk starts when AI creates deception.

Your campaign should not use AI to fake speeches, clone voices without disclosure, create false endorsements, show events that never happened, or mislead voters. The safest operating rule is simple:

“Use AI to improve campaign production. Do not use AI to fake political reality.”

The Election Commission of India warned political parties in May 2024 against AI-based tools to create deepfake audio or video that distorts information or spreads misinformation. It also directed parties to avoid false or misleading information, derogatory content against women, the use of children in campaigns, and depictions of violence or harm to animals.

Legal Review Note

You should treat AI-generated election content as a compliance issue, not only a creative issue. Before publishing synthetic voiceovers, AI leader avatars, deepfake-sparodiesarody, edited opponent clAI-generated rally visuals, or attack ads, get legal review.

Your legal review should cover:

• Election law risk

• Model Code of Conduct risk

• Defamation risk

• IT Rules and platform policy risk

• Consent for face, voice, and likeness

• Disclosure language

• Whether voters can mistake the content for real media

Use this test before publication:

“Can an ordinary voter mistake this content for a real statement, real event, real endorsement, or real recording?”

If the answer is yes, add a clear label, revise the content, or do not publish it.

The Core Compliance Rule

India does not have a single election-only AI-generated campaign content. You need to read the ECI directions, the IT Act, the IT Rules, the Model Code of Conduct, the Representation of the People Acts, criminal law, court directions, and platform policies together.

For daily campaign work, use this rule:

• AI for editing, formatting, translation, subtitles, accessibility, and design support is lower risk

• AI that imitates real people, creates fake facts, alters speeches, or hides synthetic media is high risk

• AI-generated or significantly altered political content needs disclosure when voters can mistake it for real content

• AI content involving opponents, caste, religion, public order, voting process, women, children, or criminal allegations needs legal review

The ECI’s 16 January 2025 advisory advised political parties to label synthetically generated campaign content. The advisory refers to labels such as “AI Generated,” “Digitally Enhanced,“ and “Synthetic Content.”

What the ECI Requires From Political Parties

The ECI’s May 2024 guidelines clearly specify the responsibilities of political parties regarding their ties during elections. Parties must avoid deepfake audio and video, misinformation, false or misleading information, derogatory content against women, the use of children in campaigns, and depictions of violence or harm to animals.

The ECI also directed parties to remove fake content within three hours of it coming to their notice, if it is under their control. Parties must warn the responsible person within the party, report unlawful information and fake accounts to platforms, and escalate unresolved issues to the Grievance Appellate Committee under Rule 3A of the IT Rules.

For you as a campaign manager, this means you need:

• A monitoring team

• A takedown process

• A reporting process

• An internal warning system

• A record of every removal

• A legal escalation path

• A platform complaint workflow

You cannot wait for a viral crisis before creating these controls.

What MeitY Rules Mean for Campaign Managers

MeitY regulates the platform and intermediary side of online harms. Its advisories and IT Rules matter because deepfakes and AI-generated misinformation spread through social media platforms, messaging apps, video platforms, and hosting services.

In December 2023, MeitY said platforms should identify and remove misinformation, false or misleading content, and material impersonating others, including deepfakes.

For your campaign, this matters when you report deepfakes or fake AI content to platforms. Your complaint should clearly explain whether the content involves:

• Deepfake audio or video

• Impersonation

• False or misleading information

• Manipulated media

• Fake account activity

• Election misinformation

• Defamation

• Hate speech

• Public order risk

Do not send weak complaints. Send links, screenshots, screen recordings, account handles, timestamps, and a short explanation of voter harm.

What Needs Clear Disclosure

Your campaign should disclose AI use when the content is synthetic, significantly altered, or likely to be mistaken for real media.

Disclose AI use for:

• AI-generated leader videos

• AI-generated candidate images

• Synthetic voiceovers

• AI cloned voices

• AI translated speeches using synthetic audio

• AI-generated rally visuals

• AI-generated supporter testimonials

• AI-generated historical recreations

• AI-generated parody using real political figures

• Digitally altered campaign ads

• AI-generated visuals used in paid political advertising

Use direct labels:

• “AI Generated.”

• “AI Generated Voice.”

• “Synthetic Content.”

• “Digitally Enhanced.”

• “Digitally Altered Video.”

• “Political Parody, Not a Real Recording.”

Do not hide the label in small text. Put it where voters notice it.

AI Voice Compliance

AI-generated political leaders’ voices are among the highest-risk content formats. A voice message feels personal and authentic, especially on WhatsApp, in IVR calls, in short videos, and in local campaign groups.

If your campaign uses a synthetic voice that sounds like a real leader, clearly disclose it.

Use this rule:

“If voters hear a real leader’s voice, they must know whether it is a real recording or an AI generated voice.”

For your own candidate’s AI voice, use:

“This ad uses an AI generated voice approved by the candidate.”

For a generic AI voiceover, use:

“Voiceover generated using AI.”

For parody, use:

“Satirical AI generated voice. Not a real recording.”

Do not clone an opponent’s voice. Do not create fake speeches. Do not create fake phone calls, fake confessions, fake apology clips, or fake withdrawal announcements.

What AI Use Is Lower Risk

AI tools are lower risk when they help your team produce accurate, approved, and non-deceptive content.

Lower risk uses include:

• Grammar correction

• Caption drafting

• Translation drafts

• Subtitle generation

• Speech outline support

• Press note drafts

• Poster layout support

• Image resizing

• Background cleanup

• Video trimming

• Audio noise reduction

• Accessibility captions

• Social media format adaptation

• Comment summarization

• Issue monitoring

• Volunteer training drafts

Even lower-risk content needs human review. AI tools can invent facts, misstate dates, fabricate quotes, cite false legal references, and make inaccurate claims.

What AI Use Needs Approval and Labels

Some AI use cases are not automatically banned; they require approval, consent, labels, and legal review.

Treat these as high-review content:

• AI-generated leader avatars

• Synthetic leader voiceovers

• AI-generated campaign videos

• AI translated speeches in synthetic voice

• Digitally recreated historical scenes

• AI-generated testimonials

• Parody using real leader likeness

• AI-enhanced rally visuals

• AI-generated attack visuals

• AI-generated images used in paid ads

Before publishing, ask:

• Is the content clearly labeled?

• Do we have consent?

• Did legal review approve it?

• Can voters mistake it for real?

• Does it involve an opponent?

• Does it include sensitive claims?

• Did we save source files and approvals?

If your team cannot answer these questions, do not publish.

What AI Use Campaigns Should Avoid

Some AI formats create direct compliance risk. Avoid them.

Do not use AI to create:

• Fake speeches by real leaders

• Opponent voice cloning

• Fake scandal videos

• Fake bribery clips

• Fake communal statements

• Fake caste-based statements

• Fake endorsements

• Fake candidate withdrawal notices

• Fake voting instructions

• Fake ECI documents

• Fake court or police records

• Fake news screenshots

• Fake rally footage presented as real

• Fake crowd visuals presented as real

• Deepfake attack ads

These formats mislead voters and expose your campaign to election complaints, platform removals, legal action, and public backlash.

Compliance Workflow Before Publishing

Every campaign should create an AI content approval workflow. Do not allow local teams, volunteers, meme pages, or district handles to publish synthetic content without review.

Your workflow should include:

• Identify whether AI was used

• Identify the tool used

• Identify whether the content shows a real person

• Check whether the content uses a real or cloned voice

• Check whether the content claims to show a real event

• Check whether the content attacks an opponent

• Check whether the content mentions voting, caste, religion, gender, crime, corruption, security, or violence

• Check whether disclosure is needed

• Confirm consent for face, voice, and likeness use

• Save the prompt, script, source files, and final asset

• Get legal review where needed

• Publish only from approved handles

This workflow gives your campaign control before content becomes a problem.

AI Content Register

Your campaign should maintain an AI content register. This protects you when a platform, journalist, opponent, lawyer, or election authority questions a campaign asset.

Record:

• Content title

• Tool used

• Content format

• Prompt or creative brief

• Source material

• Whether it uses a real person’s face

• Whether it uses a real person’s voice

• Consent record

• Disclosure label used

• Legal reviewer name, if applicable

• Approver name

• Date of publication

• Platform published on

• Final content link

• Archive copy

This register also helps you audit local teams.

Internal Rules for IT Cells

Your IT cell needs clear content rules. Do not rely on verbal instructions.

Set these internal rules:

• No AI voice cloning without written approval

• No opponent voice cloning

• No synthetic video without legal review

• No fake news style creatives

• No fake documents

• No fake endorsements

• No deepfake parody without a visible label

• No local team publishing AI content without approval

• No forwarding fake content in campaign groups

• No deletion of evidence before saving proof

• No AI content involving caste, religion, public order, voting process, women, or children without review

Add one simple instruction to every digital group:

“Do not create, post, or forward AI-generated political content unless the campaign has approved it.”

Deepfake Response Compliance

Compliance is not only about your own content. It also covers how you respond when someone targets your campaign with fake AI content.

Your response should follow this sequence:

• Save the original URL

• Take screenshots

• Record the video or audio

• Capture the account handle and timestamp

• Verify whether it is fake, edited, old, or AI-generated

• Report it to the platform

• Escalate to the election authorities when it affects voters

• Publish a verified clarification

• Ask supporters not to forward the fake

• Brief journalists and local teams

• Track reuploads

Use this public line:

“This video is fake and manipulated. The candidate did not make this statement. We have reported it to the platform and election authorities. Please do not forward it.”

Platform Policy Compliance

Your campaign must follow Indian law and platform rules. Meta, YouTube, X, and other platforms have policies on manipulated media, synthetic content, political ads, misinformation, impersonation, and harmful content.

Do not assume platform approval means your content is safe. During India’s 2024 election, The Guardian reported that Meta approved 14 of 22 AI-manipulated political ads submitted as a test by civil society groups. The report said the ads contained hate speech and misinformation, and that researchers removed them before publication.

Use the platform review as a single layer. Keep your own compliance review.

Pre-Certification and Paid Ads

Political campaign ads often need extra care because paid distribution gives content more reach and creates a record. If an AI-generated asset appears in paid advertising, your team should review it more closely.

Before running AI-generated paid ads, check:

• Does the platform require political ad authorization?

• Does the ad need disclosure for synthetic content?

• Does it need ECI or Media Certification and Monitoring Committee review?

• Does the ad use a leader’s face or voice?

• Does the ad include claims about an opponent?

• Does the ad use parody or edited footage?

• Does the ad include voting process information?

• Did legal review clear it?

If the ad uses synthetic voice, AI-generated leader visuals, attack content, or digitally altered footage, treat it as high risk.

Sensitive Content Areas

Some content areas require stricter review even when AI use seems minor.

Always review AI content involving:

• Religion

• Caste

• Tribe

• Language identity

• Regional identity

• Women

• Children

• Violence

• Public order

• National security

• Corruption allegations

• Criminal allegations

• Voting dates

• Polling booths

• EVMs

• Candidate withdrawal

• Alliances

• Endorsements

AI mistakes in these areas can create legal and political damage.

Training for Campaign Teams

Your campaign should train every content team, IT cell unit, district admin, volunteer coordinator, and influencer partner.

Training should cover:

• What counts as AI-generated content

• What counts as a deepfake

• What needs disclosure

• What content is banned internally

• How to report suspicious videos

• How to avoid forwarding fake content

• How to use approved templates

• How to save proof

• When to call legal review

• What to do in the first six hours after a deepfake appears

Give teams one rule they can remember:

“If it looks real but AI helped create it, check, label, and approve it before posting.”

Conclusion

Deepfakes and AI-generated political content have become a serious campaign compliance issue in India. The main risk is not the use of AI itself. The risk comes from deception. When AI makes voters believe that a real leader said something, attended an event, gave an endorsement, withdrew from an election, or made a harmful statement, even though none of that happened, it becomes a legal, political, and reputational threat.

Indian campaigns do not operate under a single AI election law. They must follow a combined framework that includes the Election Commission of India’s directions, MeitY advisories, the IT Act, the IT Rules, the Model Code of Conduct, the Representation of the People Acts, criminal law, platform policies, and court directions. This makes compliance more complex, but the practical rule is clear:

“Use AI for production support. Do not use AI to fake political reality.”

The Election Commission’s 2024 direction gave political parties a clear warning against deepfake audio, deepfake video, false information, misleading content, and AI-based misinformation. It also created a strong operational duty: when fake content comes to a party’s attention and falls under its control, the party must remove it within 3 hours, warn the responsible person, report fake accounts or unlawful content to the platforms, and escalate unresolved issues through the proper IT Rules route.

For campaign managers, this means AI content cannot be treated solely as a design or editing matter. Every campaign needs a compliance workflow. Before publishing AI-generated images, videos, voices, avatars, translated speeches, or parody content, the team must check whether voters could mistake it for real content. If the answer is yes, the content needs clear labeling, approval, and legal review.

AI tools are safer when they support normal campaign production. These include grammar correction, translation drafts, subtitles, caption writing, poster layout, video resizing, audio cleanup, accessibility captions, and issue monitoring. These uses improve speed and output quality without misleading voters.

AI tools become high risk when they clone voices, create fake speeches, generate fake endorsements, alter opponent videos, show fake crowds, create fake documents, or present synthetic scenes as real events. These formats can trigger election complaints, platform takedowns, legal action, and public backlash.

Political leader voice cloning needs special caution. If a campaign uses an AI-generated voice that sounds like a real leader, it should clearly disclose that. If voters hear a leader’s voice, they must know whether it is a real recording or an AI-generated voice. Consent, script approval, disclosure, and source records are essential.

Satire and parody are not automatically illegal, but they must remain clearly fictional. The dividing line is voter deception. A cartoon-style parody with a visible label carries a lower risk. A realistic cloned video of an opponent making a fake statement poses a high risk, even if the campaign later calls it humor.

Platforms such as Meta, YouTube, and X have policies on manipulated media, political ads, AI labels, impersonation, and misinformation. But platform action is not always fast enough. Campaigns should not wait for removal. They must save proof, report the content, publish a verified clarification, brief journalists, alert local teams, and track reuploads.

IT cells need a full response system. They should monitor keywords, local pages, WhatsApp alerts, Telegram channels, short video platforms, and influencer networks. They should also maintain a verified archive of authentic speeches, videos, photos, schedules, and statements to compare suspicious content quickly.

Deepfakes and AI-Generated Content in Indian Elections: FAQs

What Is AI-Generated Political Content?

AI-generated political content is campaign material created or significantly altered using artificial intelligence tools. It can include text, images, videos, voiceovers, subtitles, translations, avatars, posters, memes, robocalls, and social media ads. The risk depends on whether the content helps production or misleads voters.

Are AI Tools Banned in Indian Election Campaigns?

No. AI tools are not banned. Campaigns can use AI for grammar correction, caption writing, translation, subtitles, design, video resizing, audio cleanup, accessibility captions, and monitoring. The risk starts when AI creates fake speeches, cloned voices, false endorsements, fake events, or misleading visuals.

What Is the Safest Rule for Using AI in Election Content?

Use this rule:

“Use AI for production support. Do not use AI to fake political reality.”

If AI helps your team format, edit, translate, or summarize content, it is lower risk. If AI makes voters believe something false happened, it is a high risk.

What Did the ECI Say About AI-Based Deepfakes in 2024?

The Election Commission of India warned political parties against using AI-based tools to create deepfake audio or video that distorts information or spreads misinformation. It also directed parties to avoid false or misleading information, derogatory content against women, the use of children in campaigns, and depictions of violence or harm to animals.

What Is the Three-Hour Removal Rule?

The ECI directed political parties to remove fake content within three hours of it coming to their notice, when the content is under their control. Parties must also warn the responsible person, report unlawful content and fake accounts to platforms, and escalate unresolved issues through the proper IT Rules route.

Does AI-Generated Campaign Content Need Disclosure?

Yes, when voters can mistake it for real content. If your campaign uses AI-generated or significantly altered images, videos, voices, avatars, or synthetic campaign ads, add a clear label such as “AI Generated,” “Synthetic Content,” “Digitally Enhanced,“ or “AI Generated Voice.”

Does an AI-Generated Political Leader Voice Need Disclosure?

Yes. If the voice sounds like a real leader, candidate, opponent, celebrity, journalist, or public figure, disclose that it is synthetic. Use simple wording such as:

“This ad uses an AI-generated voice approved by the campaign.”

Can Campaigns Clone Their Own Candidate’s Voice?

A campaign should use the candidate’s AI-generated voice only with written consent, approved scripts, clear disclosure, and source records. The safer format is to disclose at the start of the audio or in the video caption that the voice is AI-generated.

Can Campaigns Clone an Opponent’s Voice for Satire?

No. Cloning an opponent’s voice poses significant legal and reputational risks. Even if the campaign calls it satire, voters can mistake it for real audio. Use a fictional narrator, cartoon style, or clear commentary instead.

What Is the Difference Between a Deepfake and Satire?

A deepfake makes fake content look real. Satire uses humor or exaggeration to criticize politics. The dividing line is voter deception. If voters can mistake the content for a real speech, real event, real recording, or real endorsement, treat it as a deepfake risk.

Is Political Parody Allowed?

Political parody is allowed when it is clearly fictional, labeled as such, non-deceptive, and not presented as real evidence. Risk increases when parody uses realistic face swaps, cloned voices, fake news formats, fake documents, or false factual claims.

What AI Content Should Campaigns Avoid Completely?

Campaigns should avoid fake speeches, cloned opponent voices, fake scandal videos, fake endorsements, fake ECI notices, fake police or court documents, fake voting instructions, fake withdrawal announcements, fake communal statements, and synthetic attack ads presented as real evidence.

What AI Uses Are Lower Risk for Campaign Teams?

Lower-risk uses include grammar correction, translation drafts, subtitle generation, caption writing, speech outlines, poster layouts, image resizing, background cleanup, video trimming, audio noise reduction, accessibility captions, issue monitoring, and volunteer training drafts.

What Should an IT Cell Do When It Finds a Deepfake Video?

The IT cell should save the URL, take screenshots, record the video, capture the account handle and timestamp, verify the content, report it to the platform, escalate it to election authorities where needed, publish a verified clarification, and track reuploads.

Should Campaigns Repost the Deepfake While Debunking It?

No. Reposting the full deepfake can help it spread. Use a blurred screenshot, a warning label, or a short-cropped frame only when needed. Share the official clarification and original proof instead.

How Should Campaigns Respond in the First Six Hours?

Campaigns should save proof in the first 30 minutes, verify the content within the first hour, report it to platforms within two hours, publish a clarification within three hours, brief journalists and local teams within four hours, and track reuploads through the sixth hour.

How Fast Do Meta, YouTube, and X Remove Political Deepfakes?

There is no reliable public average time for removing political deepfakes in India. Platforms can act quickly when complaints are clear and escalated, but enforcement varies. Campaigns should not wait for platforms. They should report the content and publish a correction quickly.

What Should a Strong Platform Complaint Include?

A strong complaint should include the post URL, screenshots, a screen recording, the account handle, the date and time of discovery, an explanation of the manipulation, source material (if available), the candidate or party affected, the election harm caused, and a request for urgent action.

Why Should Campaigns Maintain an AI Content Register?

An AI content register helps campaigns demonstrate how content was created, who approved it, whether disclosure was used, and whether consent was obtained. It should record the tool used, source files, prompts, script, consent, label, reviewer, approval person, publication date, and final asset link.

What Is the Final Compliance Rule for Campaign Managers?

Use this final rule:

“If AI helps production, review it. If AI changes reality, label it. If AI creates deception, reject it.”

This protects voter trust, candidate credibility, and campaign legal safety.