India Mandates 3-Hour Takedown for Deepfakes
India's 2026 IT Rules now mandate that platforms take down flagged AI-generated content and deepfakes within three hours. The amendment also heightens compliance requirements for content provenance and labeling. This follows moves by OpenAI and Google to introduce stricter controls to curb abusive AI imagery.
- This new 3-hour takedown window is a dramatic reduction from the previous 36-hour timeframe established under the 2021 IT Rules. For particularly sensitive material, such as non-consensual deepfake nudity, the compliance timeline is even shorter, at just two hours. - The amendment, formally known as the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026, was notified on February 10, 2026, and went into effect on February 20, 2026. - This regulatory action was spurred by growing concerns over the misuse of AI, highlighted by a viral manipulated video of actress Rashmika Mandanna and public statements from Prime Minister Narendra Modi on the dangers of deepfakes. - The rules introduce a formal definition for "synthetically generated information," covering any audio, visual, or audio-visual content created or altered by a computer that appears authentic. However, standard edits like color correction or noise reduction are exempt. - Beyond takedowns, the rules mandate that platforms require users to declare if their content is AI-generated and must embed non-alterable metadata or unique identifiers to trace the content's origin. - Non-compliance with these rules can lead to significant penalties, including the loss of "safe harbor" immunity, which protects platforms from liability for third-party content. This could expose companies to legal action under the Indian Penal Code and the IT Act. - India's 3-hour mandate represents one of the most aggressive takedown timelines globally. For comparison, the proposed TAKE IT DOWN Act in the US suggests a 48-hour window for intimate deepfakes, while the EU's AI Act, expected to be fully in effect in August 2026, focuses more on labeling requirements with a longer implementation period. - Platforms are now also required to use automated tools to proactively detect and block the circulation of illegal AI-generated content, including child sexual abuse material and content designed to impersonate others or spread misinformation.
