India orders social media platforms to take down deepfakes faster

India has ordered societal media platforms to measurement up policing of deepfakes and different AI-generated impersonations, portion sharply shortening the clip they person to comply with takedown orders. It’s a determination that could reshape however planetary tech firms mean contented successful 1 of the world’s largest and fastest increasing marketplace for net services.

The changes, published (PDF) connected Tuesday arsenic amendments to India’s 2021 IT Rules, bring deepfakes nether a ceremonial regulatory framework, mandating the labelling and traceability of synthetic audio and ocular content, portion besides slashing compliance timelines for platforms, including a three-hour deadline for authoritative takedown orders and a two-hour model for definite urgent idiosyncratic complaints.

India’s value arsenic a integer marketplace amplifies the interaction of the caller rules. With implicit a cardinal net users and a predominantly young population, the South Asian federation is simply a captious marketplace for platforms similar Meta and YouTube, making it apt that compliance measures adopted successful India volition power planetary merchandise and moderation practices.

Under the amended rules, societal media platforms that let users to upload oregon stock audio-visual contented indispensable necessitate disclosures connected whether worldly is synthetically generated, deploy tools to verify those claims, and guarantee that deepfakes are intelligibly labelled and embedded with traceable provenance data.

Certain categories of synthetic contented — including deceptive impersonations, non-consensual intimate imagery, and worldly linked to superior crimes — are barred outright successful the rules. Non-compliance, peculiarly successful cases flagged by authorities oregon users, tin exposure companies to greater ineligible liability by jeopardising their safe-harbour protections nether Indian law.

The rules thin heavy connected automated systems to conscionable those obligations. Platforms are expected to deploy method tools to verify idiosyncratic disclosures, identify, and statement deepfakes, and forestall the instauration oregon sharing of prohibited synthetic contented successful the archetypal place.

“The amended IT Rules people a much calibrated attack to regulating AI-generated deepfakes,” said Rohit Kumar, founding spouse astatine New Delhi-based argumentation consulting steadfast The Quantum Hub. “The importantly compressed grievance timelines — specified arsenic the two- to three-hour takedown windows — volition materially rise compliance burdens and merit adjacent scrutiny, peculiarly fixed that non-compliance is linked to the nonaccomplishment of harmless harbour protections.”

Aprajita Rana, a spouse astatine AZB & Partners, a starring Indian firm instrumentality firm, said the rules present absorption connected AI-generated audio-visual contented alternatively than each online information, portion carving retired exceptions for routine, cosmetic oregon efficiency-related uses of AI. However, she cautioned that the request for intermediaries to region contented wrong 3 hours erstwhile they go alert of it departs from established free-speech principles.

“The law, however, continues to necessitate intermediaries to region contented upon being alert oregon receiving existent knowledge, that excessively wrong 3 hours,” Rana said, adding that the labelling requirements would use crossed formats to curb the dispersed of kid intersexual maltreatment worldly and deceptive content.

New Delhi-based integer advocacy radical Internet Freedom Foundation said the rules hazard accelerating censorship by drastically compressing takedown timelines, leaving small scope for quality reappraisal and pushing platforms toward automated over-removal. In a connection posted connected X, the radical besides raised concerns astir the enlargement of prohibited contented categories and provisions that let platforms to disclose the identities of users to backstage complainants without judicial oversight.

“These impossibly abbreviated timelines destruct immoderate meaningful quality review,” the radical said, informing that the changes could undermine free-speech protections and owed process.

Two manufacture sources told TechCrunch that the amendments followed a constricted consultation process, with lone a constrictive acceptable of suggestions reflected successful the last rules. While the Indian authorities appears to person taken connected committee proposals to constrictive the scope of accusation covered — focusing connected AI-generated audio-visual contented alternatively than each online worldly — different recommendations were not adopted. The standard of changes betwixt the draught and last rules warranted different circular of consultation to springiness companies clearer guidance connected compliance expectations, the sources said.

Government takedown powers person already been a point of contention successful India. Social media platforms and civil-society groups person agelong criticized the breadth and opacity of contented removal orders, and adjacent Elon Musk’s X challenged New Delhi successful court implicit directives to artifact oregon region posts, arguing that they amounted to overreach and lacked capable safeguards.

Meta, Google, Snap, X, and the Indian IT ministry did not respond to requests for comments.

The latest changes travel conscionable months aft the Indian government, successful October 2025, reduced the fig of officials authorized to bid contented removals from the net successful effect to a legal situation by X implicit the scope and transparency of takedown powers.

The amended rules volition travel into effect connected February 20, giving platforms small clip to set compliance systems. The rollout coincides with India’s hosting of the AI Impact Summit successful New Delhi from February 16 to 20, which is expected to gully senior planetary exertion executives and policymakers to the country.