1. What is a Political Deepfake?

A political deepfake is synthetic or manipulated media that falsely makes it appear that a political candidate, party, public official, journalist, voter, or other person:

• said something they never said;
• did something they never did;
• endorsed a candidate or policy;
• withdrew from an election;
• committed a crime;
• insulted a community;
• announced false voting information;
• conceded or claimed an election;
• or otherwise communicated something that can affect electoral behaviour.

For example:

A candidate's face and voice are digitally generated to make a 30-second video in which the candidate allegedly announces withdrawal from the election.

That is considerably different from an ordinary false political statement.

The deepfake creates a false appearance of authenticity.

This distinction is legally important because the harm is not simply that somebody has expressed an incorrect opinion; it is that technology may manufacture apparently authentic evidence of something that never occurred.

2. Why Deepfakes Are Particularly Dangerous During Elections

Deepfakes can interfere with elections in at least six ways.

1. Candidate impersonation

A candidate can apparently announce:

"I am withdrawing my candidature."

2. False endorsements

A candidate can apparently endorse another political party.

3. Voter suppression

A fake government or ECI announcement could falsely state:

"Polling has been postponed."

or:

"Voters in this constituency should vote tomorrow."

4. Communal or social incitement

A fabricated speech can make a candidate appear to insult a religious, linguistic, caste or regional community.

5. Character assassination

Synthetic audio/video can falsely depict a candidate accepting bribes or making obscene statements.

6. Election-result manipulation

Fake videos can falsely claim:

"The ECI has cancelled the election."

or:

"The candidate has already conceded."

The ECI itself recognized in 2024 that AI-generated deepfakes could distort information, spread misinformation and threaten the integrity of the electoral process.

3. Constitutional Framework

A. Article 19(1)(a): Freedom of Speech

Political expression lies at the heart of Article 19(1)(a).

Therefore, a deepfake regulatory framework cannot simply prohibit everything that is:

"false political speech."

There must be a distinction between:

political criticism
and
deceptive synthetic impersonation.

For example:

"I believe Candidate X is incompetent."

is ordinary political opinion.

But:

AI-generated video falsely depicting Candidate X saying "I have accepted ₹10 crore to withdraw from the election"

creates a very different legal problem.

Restrictions must therefore be examined under Article 19(2) and must satisfy constitutional requirements such as legality, legitimate purpose and proportionality.

4. Shreya Singhal v Union of India

(2015) 5 SCC 1

This is the foundational Indian case concerning online speech.

The Supreme Court struck down Section 66A of the Information Technology Act, 2000 as unconstitutional.

The Court distinguished between:

1. discussion;
2. advocacy; and
3. incitement.

The judgment is highly relevant to deepfake regulation because the State cannot suppress protected political speech merely because it is controversial, offensive or unpopular.

Application to deepfakes

A regulatory system should therefore focus on objectively harmful conduct such as:

• impersonation;
• fraud;
• incitement;
• unlawful threats;
• electoral deception;
• privacy violations;
• unlawful misinformation;
• manipulation designed to cause legally cognizable harm.

It should not become a mechanism for suppressing legitimate criticism of candidates.

5. S. Rangarajan v P. Jagjivan Ram

(1989) 2 SCC 574

The Supreme Court held that freedom of expression cannot be restricted merely because some people may react adversely to it.

The State must demonstrate a sufficiently proximate connection between the expression and the threatened harm.

Relevance

This becomes important where authorities seek to remove political deepfakes.

The legal question should be:

What actual or legally recognized harm does the synthetic content create, and how closely is the content connected to that harm?

A fabricated video designed to trigger communal violence is fundamentally different from a clearly satirical AI parody.

6. Free and Fair Elections as a Constitutional Principle

Indian constitutional jurisprudence recognizes free and fair elections as central to constitutional democracy.

Deepfakes can threaten this principle because elections depend upon voters being able to make choices based on reasonably reliable information.

The Supreme Court has repeatedly connected electoral participation with freedom of expression.

In election-related jurisprudence, the Court has recognized that voters have an important right to information about candidates.

For example, the Supreme Court has reaffirmed that information about candidates is connected with the voter's Article 19(1)(a) rights and democratic choice.

This produces an interesting constitutional equation:

Voter's right to receive information + candidate's right to political expression + State's interest in free and fair elections

Deepfake regulation has to balance all three.

7. Union of India v Association for Democratic Reforms

(2002) 5 SCC 294

This is one of India's most important electoral-transparency decisions.

The Supreme Court recognized the voter's right to obtain relevant information concerning candidates.

The principle is highly relevant to AI-generated election misinformation.

If voters have a constitutional interest in receiving truthful and material information, deliberate synthetic fabrication designed to create a false impression about a candidate can undermine the informational foundation of democratic choice.

The later Supreme Court jurisprudence has continued to recognize the connection between candidate information, voter choice and Article 19(1)(a).

8. PUCL v Union of India

(2003) 4 SCC 399

The Supreme Court further developed the voter's right to know.

The decision emphasized the importance of information to the electoral process.

Deepfake implication

A democratic election assumes that voters are making an informed choice.

Deepfakes attack that assumption by creating fabricated information disguised as authentic information.

Thus, a strong safeguard should protect not merely the physical integrity of the ballot but also the informational integrity of the electoral environment.

9. Election Commission of India: 6 May 2024 Advisory

This is one of the most directly relevant Indian measures.

During the 2024 Lok Sabha election, the ECI directed political parties and their representatives to use social media responsibly and ethically.

The Commission specifically warned against:

• deepfakes;
• AI-generated distorted content;
• misinformation;
• disinformation;
• patently false or misleading information.

It directed political parties to remove such content within three hours of it being brought to their notice and to take action against the responsible person within the party.

The ECI specifically connected the issue with:

• the IT Act;
• IT Rules, 2021;
• IPC provisions then applicable;
• Representation of the People Acts, 1950 and 1951;
• Model Code of Conduct.

This is important because it shows that deepfake election interference was not treated as a completely new legal category. Rather, the ECI attempted to apply the existing election and technology framework to emerging AI harms.

10. ECI's January 2025 AI Advisory

The ECI subsequently issued an advisory dated 16 January 2025 concerning synthetic/AI-generated content used by political parties during election campaigning.

It emphasized transparency and accountability concerning AI-generated:

• images;
• videos;
• audio.

The advisory expressly referred back to the 2024 election guidelines and the Delhi High Court proceedings concerning deepfakes.

Basic safeguard

Political parties using synthetic content should make its AI-generated nature clear rather than presenting it as genuine footage.

This establishes an important principle:

Disclosure is preferable to allowing voters to mistake synthetic media for authentic political communication.

11. The New IT Rules Framework for Synthetic Media

India's legal position has developed further.

MeitY's 2026 amendments to the IT Rules, 2021 specifically address synthetically generated information (SGI).

The updated IT Rules were published with the February 2026 amendments and corrigenda.

The government's explanatory materials describe SGI as artificially or algorithmically created, generated, modified or altered audio-visual information made to appear real/authentic/true in circumstances where it is likely to be perceived as indistinguishable from a natural person or real-world event.

This is significant because the regulatory framework now specifically recognizes the technological phenomenon underlying deepfakes.

12. Labelling as the First Safeguard

A comprehensive election framework should require:

CLEAR AI/SYNTHETIC CONTENT LABEL

The purpose is not necessarily to prohibit synthetic political expression.

Instead, it prevents deception about authenticity.

For example:

Acceptable

"AI-generated parody video — not an actual statement by Candidate X."

Problematic

Video presented as authentic campaign footage even though the candidate never made the statement.

This distinction is constitutionally valuable because it preserves political creativity and satire while reducing deception.

13. Chaitanya Rohilla v Union of India

Delhi High Court — 21 November 2024

This is one of the most important recent Indian judicial developments concerning deepfakes.

The Delhi High Court was presented with concerns about the increasing harm caused by deepfakes and the delay between:

creation → detection → removal.

The Court found merit in the petitioners' submissions and directed the Union Government to constitute/nominates members to a committee dealing with deepfakes.

The Court specifically directed the committee to consider:

• foreign regulatory frameworks;
• technology platforms;
• telecommunications providers;
• victims;
• websites that create/deploy deepfakes;
• stakeholder experiences and suggestions.

It directed the committee to submit its report expeditiously.

Legal significance

The case demonstrates judicial recognition that deepfakes cannot be addressed solely through conventional content-moderation mechanisms.

A multi-stakeholder regulatory architecture is necessary.

14. Deepfake Detection Cannot Be the Only Safeguard

A major mistake would be to rely exclusively on AI detection tools.

Why?

Because detection technology itself can become outdated.

Synthetic media can be:

• compressed;
• cropped;
• re-encoded;
• translated;
• altered repeatedly;
• combined with genuine footage;
• generated by newer models.

Therefore:

Detection should be one layer of protection, not the entire system.

A better model is:

Detection + provenance + labelling + rapid response + human verification + platform accountability + voter education.

15. Provenance and Content Credentials

A strong technical safeguard is content provenance.

The system should preserve information showing:

• who created the media;
• when it was created;
• whether AI was used;
• what modifications occurred;
• which software generated it;
• whether the file has been subsequently altered.

This is particularly useful during elections because a viral clip can be verified through its provenance rather than relying solely on visual inspection.

16. Rapid Election-Time Takedown

Time is crucial.

Suppose:

A fake video is uploaded 12 hours before polling.

Even if it is removed 48 hours later, the electoral harm may already have occurred.

Therefore, election-specific procedures should provide:

Tier 1 — Emergency content

Candidate impersonation + voting instructions + incitement.

Immediate human review.

Tier 2 — Serious deceptive political content

Fake endorsement or fabricated statement.

Priority review.

Tier 3 — Satire/parody

Clearly identified synthetic political expression.

Normally protected unless independently unlawful.

The ECI's 2024 three-hour removal direction for political parties illustrates this election-sensitive approach.

17. Authentication Channels

One particularly effective safeguard is an official verification channel.

For example:

Candidate's verified account
+
ECI official communication channel
+
official party communication channel

If a purported candidate announcement suddenly circulates on WhatsApp or X, voters should be able to verify it immediately through authoritative channels.

The ECI already operates a "Myth vs Reality" mechanism to identify and counter election misinformation. Its 2024 archive contains examples involving false claims about EVMs, polling and other election-related misinformation.

This suggests an important institutional model:

Detection → verification → authoritative correction → public amplification.

18. Preservation of Evidence

Takedown should not automatically mean destruction of evidence.

For serious election-related deepfakes, authorities should preserve:

• original file;
• hash value;
• upload timestamp;
• URL;
• account information;
• metadata;
• repost history;
• advertisements associated with the content;
• engagement data;
• copies submitted by complainants.

This is essential for subsequent:

• criminal investigation;
• election petitions;
• civil proceedings;
• platform liability;
• attribution analysis.

19. Attribution of the Creator

Another difficult problem is:

Who created the deepfake?

Possible actors include:

• political parties;
• campaign consultants;
• supporters;
• foreign actors;
• anonymous users;
• bots;
• advertising agencies;
• coordinated networks.

Therefore, safeguards should focus not merely on the person appearing in the deepfake but also on the origin and distribution network.

20. Political Advertising Rules

AI-generated political advertisements should be subject to heightened transparency.

A robust system should require:

1. disclosure that AI was used;
2. identification of the sponsoring entity;
3. preservation of the original creative;
4. preservation of targeting parameters;
5. archive of the advertisement;
6. rapid complaint mechanism;
7. special restrictions immediately before polling.

The ECI has increasingly emphasized transparency around AI-generated campaign content.

21. Privacy and Personality Rights

Deepfakes also implicate Article 21.

A candidate's:

• face;
• voice;
• name;
• likeness;
• persona;

can be appropriated without consent.

Indian courts have increasingly granted protection against unauthorized exploitation of personality attributes through AI and deepfakes.

For example, the Delhi High Court has in recent cases considered claims involving unauthorized AI-generated use of a person's name, image, voice, likeness and persona.

This provides a second legal pathway:

A political deepfake may be unlawful not only because it interferes with elections, but also because it violates the personality/privacy/reputation rights of the person depicted.

22. Defamation

A deepfake falsely portraying a candidate accepting a bribe, committing a crime or making a defamatory statement may also give rise to defamation-related remedies, depending on the facts and applicable criminal/civil law.

The important distinction is:

A deepfake can manufacture the appearance of evidence.

That makes it potentially more damaging than an ordinary defamatory statement.

23. Representation of the People Act

The Representation of the People Act, 1951 is central to election regulation.

Depending on the facts, deepfake conduct may intersect with provisions concerning:

• corrupt practices;
• undue influence;
• false statements;
• election offences;
• campaign conduct.

The precise provision depends heavily on:

• who created the deepfake;
• who published it;
• whether it was connected with a candidate;
• the nature of the statement;
• the election involved;
• timing;
• intention;
• effect on the electoral process.

The ECI itself identified the RPA framework as part of the legal framework applicable to deepfake-related election conduct.

24. Model Code of Conduct

The Model Code of Conduct is especially important during an election period.

The ECI's 2024 directions explicitly tied deepfake restrictions to the MCC.

The central objective is to maintain a:

level playing field between electoral contestants.

A political deepfake can distort that level playing field by creating false campaign material that is difficult for the targeted candidate to answer before voting occurs.

25. Case Law on Election Commission's Powers

Mohinder Singh Gill v Chief Election Commissioner

(1978) 1 SCC 405

A foundational case concerning the Election Commission's constitutional powers.

The Supreme Court recognized the broad constitutional role of the ECI in ensuring the conduct of free and fair elections under Article 324.

Deepfake application

Article 324 provides an important constitutional basis for the ECI to respond to new technologies affecting election integrity where existing legislation does not completely address the problem.

However, ECI powers remain subject to:

• the Constitution;
• legislation;
• judicial review;
• fundamental rights.

26. A.C. Jose v Sivan Pillai

(1984) 2 SCC 656

The Supreme Court considered the limits of the Election Commission's power under Article 324.

The case is important because Article 324 does not mean that the ECI can disregard statutory provisions.

Application

The ECI can respond to deepfakes through its constitutional and regulatory powers, but its measures must remain within the legal framework governing elections.

This is why the combination of:

Article 324 + RPA + MCC + IT Rules

is particularly important.

27. Election Commission of India v Ashok Kumar

(2000) 8 SCC 216

The Supreme Court emphasized the ECI's role in ensuring free and fair elections while also explaining the limits on judicial intervention during the electoral process.

Relevance

A deepfake complaint arising immediately before polling presents a classic problem:

Should the matter wait for ordinary litigation?

Often, waiting defeats the remedy.

Election-specific administrative intervention may therefore be necessary, subject to legality and due process.

28. K.S. Puttaswamy v Union of India

(2017) 10 SCC 1

Although not an election-deepfake case, Puttaswamy is important for:

• informational privacy;
• autonomy;
• dignity;
• control over personal information.

A voice clone or facial clone involves use of attributes closely connected with identity.

Therefore:

Deepfake governance should protect both electoral integrity and individual dignity.

29. The Free-Speech Problem: Not Every Deepfake Should Be Banned

This is perhaps the most important constitutional safeguard.

Imagine:

A political cartoon uses AI to place a politician's face on a fictional character and clearly states that it is satire.

An absolute ban on all synthetic political material could unnecessarily restrict Article 19(1)(a).

Compare:

A fake video designed to look exactly like genuine campaign footage and falsely showing a candidate withdrawing from the election.

The second presents a much stronger case for intervention.

Therefore, the regulatory distinction should be:

Synthetic political expression

versus

deceptive synthetic impersonation.

30. Recommended Safeguard Framework

A comprehensive Indian framework could operate as follows:

31. Three-Hour Rule as a Model for Election Emergencies

The ECI's 2024 direction is particularly significant because it recognizes that ordinary platform response times may be inadequate during elections.

The logic is:

The shorter the period before voting, the greater the potential harm from delay.

Thus, election-period regulation can legitimately justify faster procedures than ordinary content moderation, provided adequate safeguards exist against arbitrary censorship.

32. Independent Verification Cell

One practical institutional solution would be an Election AI Verification Cell comprising:

• ECI officials;
• forensic experts;
• election-law experts;
• AI specialists;
• representatives of major platforms;
• cybercrime authorities.

Its role would be to determine rapidly:

1. Is the content synthetic?
2. Is the person actually depicted?
3. Is the content being presented as genuine?
4. Is it election-related?
5. Is there evidence of impersonation?
6. Does it violate election law?
7. Does it require immediate removal?
8. Should it instead receive an AI/satire label?

This is consistent with the multi-stakeholder approach contemplated by the Delhi High Court in Chaitanya Rohilla.

33. Foreign Actors and Coordinated Inauthentic Behaviour

Deepfake election interference may originate outside India.

A sophisticated campaign could combine:

AI-generated videos + fake accounts + bots + targeted advertising + encrypted messaging + coordinated reposting.

Therefore, regulation of the individual video alone may be inadequate.

Authorities should investigate the network, not merely the content.

The policy literature increasingly identifies coordinated synthetic media, bots and synthetic identities as interconnected election-interference techniques.

34. Why Human Review Remains Necessary

AI detectors can produce:

• false positives;
• false negatives;
• language bias;
• compression-related errors;
• difficulty with new generation models.

Research has found that humans themselves can struggle to identify political deepfakes, particularly sophisticated synthetic audio.

Accordingly:

AI detector result ≠ final legal determination.

A human reviewer should examine:

• provenance;
• context;
• source;
• metadata;
• surrounding posts;
• original footage;
• candidate's authentic communications.

35. Election-Specific Voter Education

A safeguard framework should not depend entirely on government censorship.

Voters should be taught:

Before sharing a political video:

STOP

CHECK

VERIFY

SHARE

For example:

"Did the candidate's official account publish this?"

"Does the ECI website confirm the alleged election announcement?"

"Is the video labelled AI-generated?"

"Does another reputable source report the same event?"

This is particularly important because fake content can spread faster than institutional correction.

The ECI's existing "Myth vs Reality" system illustrates the value of authoritative correction mechanisms.

36. The Most Important Legal Balance

Deepfake regulation should satisfy five constitutional principles:

1. Legality

There must be a clear legal basis.

2. Legitimate purpose

Protection of elections, public order, privacy, reputation or other constitutionally recognized interests.

3. Necessity

The restriction should address a genuine problem.

4. Proportionality

The restriction should not unnecessarily suppress legitimate political expression.

5. Procedural safeguards

There should be:

• notice where feasible;
• reasons for removal;
• appeal/review mechanisms;
• preservation of evidence;
• protection against political misuse of takedown powers.

37. Important Case-Law Matrix

The Delhi High Court's Chaitanya Rohilla order is particularly important because it specifically addressed the institutional problem of deepfake creation, detection and removal.

38. Proposed Legal Test for an Election Deepfake

A court or election authority could consider the following questions:

Question 1

Is the content synthetic or materially manipulated?

Question 2

Is a real candidate/public figure being impersonated?

Question 3

Is the content presented as authentic?

Question 4

Was AI manipulation disclosed?

Question 5

Is the content reasonably capable of affecting electoral choice?

Question 6

Does it contain a false statement of fact rather than protected opinion/satire?

Question 7

Does it violate an existing electoral, criminal, civil or IT-law prohibition?

Question 8

Is immediate intervention necessary because of proximity to polling?

Question 9

Would removal be proportionate, or would a label/correction be sufficient?

Question 10

Are procedural safeguards available against wrongful removal?

This produces a more constitutionally defensible framework than simply saying:

"All AI political content is prohibited."

39. Overall Legal Position in India

The Indian approach is developing through multiple overlapping legal regimes rather than one single "Deepfake Election Act."

The framework presently consists broadly of:

Constitution

• Article 19(1)(a)
• Article 19(2)
• Article 21
• Article 324

Election law

• Representation of the People Acts
• Model Code of Conduct
• ECI directions/advisories

Technology law

• Information Technology Act, 2000
• IT Rules, 2021, as amended
• 2026 synthetic-generated-information framework

Civil/private law

• privacy
• personality rights
• passing off
• defamation
• injunctions

Institutional mechanisms

• ECI fact-checking
• platform grievance systems
• cybercrime authorities
• courts.

The ECI's 2024 directions already specifically prohibited political parties from circulating deepfake audio/video and required prompt removal after notice. The 2025 ECI advisory then emphasized labelling AI-generated campaign material. The IT Rules' 2026 amendments provide a further technology-specific regulatory layer for synthetic media.

40. Conclusion

The central legal problem with election deepfakes is not simply false speech.

It is the ability of AI to manufacture apparently authentic political evidence at unprecedented speed and scale.

A constitutionally sound Indian safeguard framework should therefore combine:

AI disclosure + provenance + rapid detection + human verification + election-specific takedown + evidence preservation + authoritative correction + platform accountability + voter education + judicial remedies.

At the same time, regulation must distinguish deceptive impersonation from legitimate political criticism, parody and satire. Shreya Singhal and S. Rangarajan make clear why overbroad restrictions on political expression would raise serious Article 19 concerns.

The strongest legal foundation is therefore a combination of:

Article 19(1)(a) + Article 19(2) + Article 21 + Article 324 + RPA + MCC + IT Rules + ECI directions.

And the most directly relevant recent judicial authority is Chaitanya Rohilla v Union of India, because the Delhi High Court expressly recognized the need for a coordinated institutional response to the creation, detection and removal of deepfakes.