🏛️ Polity & GovernanceMAINS · GS2.7 · GS2.8 · GS3.18

ECI tightens rules on AI-generated poll content

The Election Commission has ordered campaign deepfakes and synthetic media to be labelled, and unlawful AI content pulled down within three hours, across the 2026 five-state assembly elections.

What happened

The Election Commission of India (ECI) issued a fresh advisory on the use of artificial-intelligence and synthetic media in campaigning for the ongoing 2026 assembly elections, insisting that all digital campaigning comply with the Information Technology Act, 2000, the IT Rules, 2021 and the Model Code of Conduct (MCC).
Parties, candidates and campaign representatives must clearly label any AI-generated or AI-altered campaign material as "AI-Generated", "Digitally Enhanced" or "Synthetic Content", and disclose the originating entity behind it.
Any misleading or unlawful AI-generated or manipulated content is to be acted upon within 3 hours of being flagged to the social-media platform.
The order covers the five legislatures going to polls — Assam, Kerala, Tamil Nadu, Puducherry and West Bengal — where State IT Nodal Officers notified under the IT Act monitor and act on MCC violations, content disrupting law and order, and false narratives against the polling process.
Since elections were announced on 15 March 2026, over 11,000 social-media posts/URLs have been identified and acted upon — through content removal, FIRs, clarifications and rebuttals.
The Commission reiterated Section 126 of the Representation of the People Act, 1951, which bars the display of election matter in polling areas during the 48-hour "silence period" before polling closes.
On grievance redress: from 15 March to 19 April, 3,23,099 complaints were filed through the C-Vigil module on ECINET, of which 3,10,393 (96.01%) were resolved within the stipulated 100-minute window.

Background & context

The body that issued this direction — the Election Commission of India — is a permanent, independent constitutional body created by Article 324 of the Constitution, charged with the superintendence, direction and control of elections to Parliament, the State legislatures, and the offices of President and Vice-President. It is not a statutory creation: it predates the statute that governs the conduct of elections. The ECI came into existence on 25 January 1950 (now observed as National Voters' Day), and from 1 October 1993 it has functioned as a multi-member body — a Chief Election Commissioner sitting with Election Commissioners, who enjoy salary and status equivalent to a judge of the Supreme Court. The CEC can be removed only by the same process as a Supreme Court judge (Parliamentary address on grounds of proved misbehaviour or incapacity); the other Commissioners can be removed only on the CEC's recommendation. The qualifications, term and conditions of service of the Commissioners are now set by the Chief Election Commissioner and Other Election Commissioners (Appointment, Conditions of Service and Term of Office) Act, 2023.

The advisory does not flow from any new statute. It is an exercise of the ECI's existing toolkit. The first leg is the Model Code of Conduct — a set of conventions agreed among political parties that the ECI enforces the moment elections are announced. The MCC is not law in the strict sense; it has no dedicated penal statute behind it, and its force comes from the ECI's constitutional authority and the parties' consent. The second leg is the statutory frame for intermediaries: the Information Technology Act, 2000 and the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021, which place due-diligence and takedown obligations on social-media platforms and let the government notify nodal officers. The three-hour takedown timeline the ECI cites mirrors the expedited removal expectations already built into the intermediary-rules framework for unlawful content. The third leg is the Representation of the People Act, 1951 — the principal statute that actually governs the conduct of elections (its companion, the RPA 1950, deals with the allocation of seats and the preparation of electoral rolls).

The trigger for this round of tightening is the arrival of cheap, convincing generative AI in the campaign. Deepfaked voices, face-swapped video, AI-cloned speeches and synthetically generated "news" can now be produced at scale and pushed across platforms faster than any manual review can catch them — and they are especially dangerous in the closing hours before a vote, when there is no time for a rebuttal to circulate. The ECI's answer is not to ban the technology but to force transparency (mandatory labelling and source disclosure) and speed (a fixed takedown clock), so that voters can tell genuine material from manufactured material and unlawful content does not stay up long enough to do damage.

For Prelims

Issuing body: Election Commission of India — a constitutional body under Article 324; established 25 January 1950; multi-member since October 1993.
Labelling mandate: AI/synthetic campaign content must carry one of three labels — "AI-Generated", "Digitally Enhanced" or "Synthetic Content" — plus disclosure of the originating entity.
Takedown clock: unlawful AI-generated or manipulated content to be acted on within 3 hours of notice to the platform.
Legal basis (3 pillars): IT Act 2000 · IT (Intermediary Guidelines) Rules 2021 · Model Code of Conduct. No new law was enacted — the direction rides on existing instruments.
Section 126, RPA 1951: prohibits the display of any election matter in polling areas during the 48-hour silence period before the close of poll (the "poll-eve" / "campaign silence" ban).
States voting: Assam, Kerala, Tamil Nadu, Puducherry and West Bengal. (Puducherry is a Union Territory with a legislative assembly.)
Enforcement front line: State IT Nodal Officers notified under the IT Act monitor MCC violations, law-and-order content and false narratives.
Scale of action: 11,000+ posts/URLs acted on since 15 March 2026 (removals, FIRs, clarifications, rebuttals).
Grievance tool — C-Vigil on ECINET: 3,23,099 complaints lodged (15 Mar–19 Apr); 96.01% resolved within the 100-minute service window. C-Vigil lets any citizen geo-tag and timestamp a photo/video of an MCC violation; ECINET is the ECI's integrated digital ecosystem of apps that now houses it.
What it is NOT: This is not a new statute or ordinance banning AI — it is an advisory/direction using existing law; it does not outlaw AI in campaigns, only mandates labelling and rapid takedown of unlawful material. The ECI is not a statutory body — it is constitutional (contrast it with, say, the National Human Rights Commission, which is statutory). Section 126 is not in the RPA 1950 (rolls/seats) but in the RPA 1951 (conduct of elections). The silence period is 48 hours, not 24.
The constitutional/electoral-law set to keep straight: Art 324 (ECI's powers) · Art 325 (one general electoral roll; no exclusion by religion, race, caste, sex) · Art 326 (universal adult suffrage, voting age 18 after the 61st Amendment, 1988) · Art 327–328 (Parliament/State legislature power to make election law) · Art 329 (bar on judicial interference in electoral matters except by election petition). Statutes: RPA 1950 (rolls & seat allocation) · RPA 1951 (conduct of elections, qualifications, disqualifications, offences, election petitions).

For UPSC: Section 126, RPA 1951 = the 48-hour "silence period" / poll-eve ban on displaying election matter in polling areas. The ECI is a constitutional body (Art 324), and its AI-labelling direction rides on the IT Rules 2021 + MCC — not on any new law. C-Vigil (on ECINET) = citizen-sourced MCC-violation reporting with a 100-minute resolution promise.

Why it matters

The problem this addresses is the collapse in the cost and the rise in the realism of fabricated political media. A single convincing deepfake of a candidate "confessing", "endorsing" or "abusing" can move votes in a close seat — and if it lands in the final 48 hours, the target has no practical way to rebut it before polling. Generative AI also lets a small operation flood many platforms at once, overwhelming the slow, complaint-by-complaint takedown model that worked when fake content was hand-made. By forcing every synthetic asset to wear a label and naming the originator, the ECI shifts part of the burden onto the maker and gives the voter a cue to discount manufactured material. The fixed three-hour clock attacks the second half of the problem — dwell time — by compressing how long unlawful content can stay live. The model-code reminder and the Section 126 citation tie the new digital rules back to the old principle that the campaign must fall silent before the vote, so that the elector decides in quiet. The C-Vigil numbers (over three lakh complaints, 96% cleared inside 100 minutes) show the ECI leaning on crowd-sourced, geo-tagged citizen reporting as the sensing layer for an electorate too large to police centrally. Taken together, the direction is an attempt to keep the integrity of the "level playing field" — the core promise of the MCC — intact in an information environment that generative AI has made far harder to police.

For Mains

Anchor

A question on how India is regulating AI/deepfakes in elections, or on the evolving powers and tools of the Election Commission, can be built directly around this ECI direction — labelling mandate, three-hour takedown, and the IT Act/IT Rules/MCC triad.

Substantiation

Hard data to quote: 11,000+ posts/URLs acted on since 15 March 2026; 3,23,099 C-Vigil complaints with 96.01% resolved inside 100 minutes — concrete evidence of the scale of digital-content moderation in a live Indian election.

Exemplification

A ready example of "soft-law" enforcement — the ECI extending the Model Code of Conduct and existing IT Rules to a new threat without waiting for fresh legislation, illustrating regulatory adaptation through executive direction.

Problematisation

The direction implicitly admits the gaps: enforcement depends on platform cooperation and on State IT Nodal Officers; the MCC has no penal teeth of its own; a three-hour window may still be slower than viral spread; and labelling is hard to enforce against anonymous or cross-border originators.

Way-forward

Points to a layered model — transparency (mandatory labelling + source disclosure), speed (fixed takedown timelines), citizen sensing (C-Vigil/ECINET), and statutory backing (IT Rules) — as a template for governing synthetic media beyond the election context.

Position

The government/ECI's stated stance: regulate the harm, not the tool — permit AI in campaigning but compel disclosure and act fast on the unlawful, preserving both free speech and electoral integrity.

Deploys into: integrity of the electoral process and the role of the Election Commission (GS2.7, RPA); statutory/constitutional bodies and their evolving mandate (GS2.8); cyber security, misinformation and the regulation of social-media intermediaries (GS3.18); ethics of technology and probity in public information.

Election Commission · 2026-04-19 · PRID 2253528 · PIB source ↗

Related: ECI & electoral-law hub · Polity & Governance · This week's cards