India Slashes AI Deepfake Takedown Window to 3 Hours, Mandates Irremovable Labels

States have tried to tame disruptive media before: India’s own Section 66A (added 2008, struck down 2015) similarly forced quick deletions of ‘offensive’ posts, while the U.S. 1934 Communications Act pushed radio networks to self-censor ‘indecent’ broadcasts. Each wave pairs a new technology with tighter timelines and liability threats, then collides with free-speech jurisprudence. The 2026 amendments extend that centuries-long tug-of-war—printers in 1790s France, telegraphs in 1860s Britain, social media in the 2010s—to generative AI. They reflect two structural trends: (1) sovereigns moving from voluntary advisories to hard-coded algorithmic compliance, and (2) shifting the cost of policing from the state to private intermediaries, effectively deputising tech firms as real-time censors. Whether this matters in 2126 will hinge on precedent: if courts uphold the three-hour rule, it may normalise near-instant, machine-led gatekeeping worldwide; if struck down, it will reiterate the cyclical limits of state control over information flows. Either way, India—home to the world’s largest user base for several global platforms—has signalled that AI governance is no longer theoretical but enforceable within hours.