India orders social media platforms to take down deepfakes faster

India has issued new directives for social media platforms, demanding they enhance their oversight of deepfakes and other AI-created impersonations. This significant policy shift comes with a stark reduction in the time allowed for compliance with takedown requests, a move poised to alter the landscape of content moderation for tech giants operating in one of the world’s most expansive internet markets. The amendments, released on Tuesday as updates to India’s 2021 IT Rules, formally incorporate deepfakes into a regulatory framework. Under these new guidelines, platforms are required to label and trace synthetic audio and visual content. Compliance timelines have been drastically shortened—social media companies now have just three hours to respond to official takedown requests and two hours for urgent user complaints. With over a billion internet users and a youthful demographic, India's digital market holds significant sway. This means that the compliance measures adopted by platforms in India could ripple outwards, influencing global moderation strategies. The new regulations stipulate that social media platforms must verify and disclose whether content is AI-generated, ensuring that deepfakes are clearly identified and include traceable provenance data. Certain types of synthetic content, such as deceptive impersonations and non-consensual intimate imagery, are strictly prohibited. Companies that fail to comply, especially in cases highlighted by authorities or users, risk losing their safe-harbour protections under Indian law, exposing them to increased legal risks. Rohit Kumar, a partner at The Quantum Hub, emphasized that these revised IT Rules represent a more nuanced approach to regulating AI-generated deepfakes. He noted that the compressed grievance timelines would significantly heighten compliance demands for social media companies, raising concerns about the implications for free speech. Aprajita Rana, a partner at AZB & Partners, pointed out that while the rules now concentrate on AI-generated audio-visual content, they also exempt certain benign uses of AI. However, she highlighted that the expectation for platforms to remove content within three hours of becoming aware of it deviates from established free speech principles. The law mandates that such removals occur swiftly, with labeling requirements designed to combat the spread of harmful content. The Internet Freedom Foundation, a digital advocacy organization based in New Delhi, cautioned that these new regulations could lead to increased censorship. They argue that the rapid takedown timelines hinder meaningful human review, pushing platforms to rely heavily on automated systems which may result in excessive content removal. Concerns have also been raised regarding the expanded list of prohibited content and the potential for platforms to share user identities with complainants without judicial oversight. The Foundation's statement on social media emphasized that these stringent timelines could erode protections for free speech and due process. In light of the amendments, industry insiders have suggested that the changes were made with limited consultation, noting that many recommendations were overlooked in the final rules. The substantial alterations from the draft to the final version warranted further discussions to clarify compliance expectations. The issue of government takedown powers has long been a contentious topic in India, with social media companies and civil society groups criticizing the vague and broad nature of content removal orders. Notably, even X, formerly known as Twitter, has legally challenged the Indian government regarding the scope and transparency of such directives. The latest amendments are set to take effect on February 20, leaving little time for platforms to adapt their compliance processes. This rollout coincides with India's hosting of the AI Impact Summit in New Delhi, drawing attention from global technology leaders and policymakers.