Election Commission of India: Advisories on Labelling and Disclosure of AI-Generated Campaign Content (16 January 2025 and 24 October 2025)

The ECI first addressed AI-generated content in its guidelines of 6 May 2024 to recognised political parties, issued during the Lok Sabha election after the Delhi High Court directed it to decide a representation on deepfakes. Those guidelines prohibited deceptive synthetic content and required deepfake audio or video on party handles to be removed within three hours. They did not require parties to disclose campaign material that was AI-generated but not deceptive.⁵

The use of AI in campaigns grew over the following months. Ahead of the November 2024 Maharashtra Assembly election, the BJP posted an audio clip attributed to NCP (SP) MP Supriya Sule that fact-checkers found to be manipulated. Sule complained to the police and the ECI. In January 2025, during the Delhi Assembly campaign, Delhi Police registered an FIR against the Aam Aadmi Party over an AI-generated video on its official X handle that swapped the faces of Prime Minister Narendra Modi and Home Minister Amit Shah onto villains from a 1990s Bollywood film.⁷ The Commission issued its first labelling advisory on 16 January 2025, days after the FIR and three weeks before Delhi voted on 5 February 2025.¹⁶

The advisory dated 16 January 2025 (No. 491/Social media-SOP/2025/Communication), titled Advisory for labelling synthetic/AI generated content used by Political Parties for election campaigning, went to the heads of all recognised national and state parties.¹ It noted that advances in AI made it possible to create highly convincing synthetic images, video and audio, and that AI-generated content could influence voter opinion and trust. It described the May 2024 guidelines as the first step of its approach. Given the deepening influence of synthetic content on public opinion, it advised parties, their leaders, candidates and star campaigners to go further and adopt measures for prominent labelling.¹

The advisory set out two measures:¹

• Labelling AI or synthetically generated content. Any image, video, audio or other material generated or significantly altered by AI must be clearly labelled with a notation such as "AI-Generated", "Digitally Enhanced" or "Synthetic Content".
• Disclosure in campaign materials. Campaign advertisements and promotional content disseminated online or on other platforms must carry disclaimers wherever synthetic content is used.

The language was hortatory: parties were "urged to adhere" to the advisory. It gave no specifications for label size or placement, created no reporting or record-keeping duty, and named no sanction for non-compliance. It applied to labelling by the parties themselves, not to platforms or AI tool providers. MediaNama noted that the industry was moving in the same direction, with Meta applying "AI info" labels and OpenAI adding provenance metadata to images from its DALL-E 3 generator. Researchers had warned that sophisticated adversaries could break watermarking.⁶

The Model Code of Conduct took effect on 6 October 2025 with the announcement of the Bihar Assembly election, held in two phases on 6 and 11 November, and by-elections to eight Assembly constituencies.⁸ On 9 October the Commission announced that the MCC applies to online and social media content. It warned parties against misusing AI tools to create deepfakes and said parties, leaders, candidates and star campaigners "shall take necessary measures" to label synthetic content prominently, including in advertisements. It added that social media was being closely watched.³

On 24 October 2025 the ECI issued a more detailed advisory, No. 4/Misc/2025/SDR/Vol.XX, signed by Principal Secretary S. B. Joshi. It is titled Advisory on responsible use and disclosure of synthetically generated information and AI-generated content during elections.² It said hyper-realistic synthetic content, including content "depicting political leaders making electorally sensitive messages", was contaminating the level playing field. It described synthetic information as a deep threat because of its ability to masquerade as the truth. Invoking its plenary power under Article 324 of the Constitution, the Commission reminded parties to comply with the IT Rules, 2021 and its earlier guidelines. In continuation of the communications of 6 May 2024 and 16 January 2025, both enclosed, it issued the following directions to parties, candidates and campaign representatives:²

1. Label format. Any synthetically generated or AI-altered image, audio or video used for campaigning must carry a clear, prominent and legible label such as "AI-Generated", "Digitally Enhanced" or "Synthetic Content". The label must cover at least 10% of the visible display area, or the first 10% of the duration for audio. For video, it must appear at the top of the screen.
2. Creator disclosure. The content must prominently disclose the name of the entity responsible for generating it, in the metadata or an accompanying caption.
3. Misrepresentation ban. No content may be published or forwarded that is unlawful and misrepresents the identity, appearance or voice of any person without consent in a manner reasonably likely to mislead or deceive voters.
4. Three-hour takedown. Any synthetic or AI-altered content, misinformation or manipulated content covered by the misrepresentation ban that is detected on official party handles must be taken down within three hours of being noticed or reported.
5. Record-keeping. Parties must keep internal records of all AI-generated campaign material, including creator details and timestamps, for verification when the ECI asks for them.

The advisory took effect immediately and applies during all general elections and by-elections until further orders. Chief Electoral Officers of all States and Union Territories were directed to pass it on to parties and field officials.²

Sources: ECI letters of 6 May 2024, 16 January 2025 and 24 October 2025.⁵¹²

The October 2025 standard came two days after MeitY published its draft amendment to the IT Rules on synthetically generated information (22 October 2025). That draft proposed labels covering at least 10% of the visual area or the first 10% of an audio clip.¹⁰ MeitY notified the final amendment on 10 February 2026, in force from 20 February 2026. It requires platforms to label permitted synthetic content and attach provenance metadata, and cuts the deadline for acting on government or court takedown intimations under Rule 3(1)(d) from 36 hours to three hours.¹¹ The ECI's rules bind campaigners, while MeitY's rules bind platforms. Together they place labelling duties at both points.

Commentators welcomed the labelling rules as a transparency measure but questioned whether they could work in practice. In its report on the October 2025 advisory, MediaNama said the directions might be difficult to implement. Parties could find workarounds because the rules apply only during elections. The rules also reach only official party accounts, not unofficial or "dummy" accounts run by party social media teams, which could flout them without consequence.⁹ Earlier discussions cited by MediaNama questioned whether watermarks and labels survive determined adversaries, and suggested fingerprinting content at its source as an alternative.⁶

The advisories rely on the MCC, which has no statutory backing, and on the ECI's general power under Article 324. Neither sets out a graded penalty for failing to label. The Commission's enforcement statements refer to "strict" action and to monitoring by state IT nodal officers, but the sources reviewed for this entry report no ECI decision penalising a party specifically for an unlabelled AI post.³⁴ Litigants have increasingly turned to the courts instead. In October 2026 a Chandigarh civil court ordered Meta to suspend three Facebook pages carrying AI-generated videos of a Punjab Congress politician ahead of the 2027 Punjab election. It was the third order from courts in the Chandigarh tricity within a year directing Meta to act against AI-manipulated political content. MediaNama attributed the trend to the IT Rules' design, under which a platform's duty to remove such content typically starts only with a court order or an authorised officer's intimation.¹²

The ECI applied the framework to the 2026 Assembly elections in Assam, Kerala, Tamil Nadu, Puducherry and West Bengal, announced on 15 March 2026. On 19 April 2026 it reiterated three points. Parties, candidates and campaign representatives must label synthetic or AI-altered campaign content and disclose the originating entity. Misleading or unlawful AI-generated or manipulated content must be acted on within three hours of being brought to platforms' notice. State IT nodal officers notified under the IT Act were monitoring content that violated the MCC or spread false narratives about the polling process.⁴ The Commission reported that since 15 March 2026 more than 11,000 posts or URLs had been acted on through removals, FIRs, clarifications or rebuttals, and that its C-Vigil app had received 3,23,099 complaints, 96.01% of them resolved within 100 minutes. These figures cover all MCC complaints, not only AI-related ones.⁴

As of 14 October 2026, the 24 October 2025 advisory remains in effect for all general and bye-elections until further orders, read together with the January 2025 advisory and the May 2024 guidelines.² Since February 2026, platforms have also had their own labelling and three-hour takedown duties under MeitY's amendment to the IT Rules.¹¹ No court has ruled on the ECI's power to impose content-labelling obligations through advisories. Upcoming tests include the early-2027 Punjab Assembly election, where courts are already dealing with disputes over AI-generated political videos.¹²