Update

In a separate Lok Sabha reply on 12 August 2026, Minister of State Jitin Prasada confirmed that platforms must remove unlawful AI-generated content within three hours of a government order, and within two hours for content involving impersonation or nudity. These are the same timelines fixed by the February 2026 IT Rules amendment, now described specifically in relation to AI-generated content such as deepfakes.

Prasada outlined the legal provisions the government considers applicable to deepfake-related harms: identity theft and impersonation offences under the IT Act, and cheating by personation and forgery offences under the Bharatiya Nyaya Sanhita, which replaced the Indian Penal Code in 2023. India does not have a standalone law specifically addressing deepfakes or synthetic media.

Prasada's reply also referred to the IndiaAI Safety Institute, describing its role in conducting safety research, developing standards, testing AI systems, and evaluating emerging risks. The Institute operates under the IndiaAI Mission's "Safe and Trusted AI" pillar, an initiative that predates this parliamentary reply; it was not newly created or announced by Prasada's statement.

Analysis

Some early coverage of this reply described the IndiaAI Safety Institute as a new announcement tied to deepfake enforcement. That framing does not match the Institute's actual history: it exists as part of a broader AI governance initiative that has been under development since before the reply, and its stated functions are research and standards-setting rather than takedown enforcement or individual case adjudication. The government has not stated that the Institute has any formal role in deciding whether specific content should be removed under the IT Rules.

Legal commentators have noted that India's approach to deepfakes relies on layering existing IT Act and BNS provisions rather than creating a dedicated statute, a choice with precedent in other jurisdictions grappling with fast-evolving synthetic media technology, since a narrowly drawn standalone law risks being overtaken by new generation techniques. Digital rights groups have separately raised concerns that the layered approach leaves gaps in liability and remedy, since no single provision squarely addresses AI-generated impersonation or non-consensual synthetic imagery, leaving affected individuals to combine criminal, IT Rules, and civil remedies that were not designed with this specific harm in mind.

The comparison to similarly named institutions abroad is instructive but imperfect. The United Kingdom's AI Safety Institute was renamed the AI Security Institute in February 2025, alongside a shift in focus toward cybersecurity risks and away from broader AI ethics questions such as bias and misinformation. India's IndiaAI Safety Institute has not undergone a comparable rebrand or refocusing, and unlike its UK and US counterparts, it does not yet sit within a jurisdiction that has enacted or is actively finalising a dedicated AI statute to which its research findings would feed directly. Whether India moves toward such legislation remains an open question that officials have discussed only in general terms.