• India shortens illegal content removal to just 3 hours on social media
• Amendments cover AI-generated material, raising fears of broad censorship

India has introduced new rules requiring social media platforms to remove illegal content within three hours of notification, down from the previous 36-hour window, marking a major tightening of existing regulations.

The changes take effect on February 20 and apply to major platforms such as Meta, YouTube, and X, including AI-generated content.

The government did not explain the reason for the shorter removal window, but critics fear it is part of a wider crackdown on digital content, potentially leading to censorship in the world’s largest democracy, with over a billion internet users.

In recent years, Indian authorities have used existing IT rules to compel platforms to remove content deemed illegal under laws concerning national security and public order, giving authorities broad powers over social media content.

Transparency reports show that over 28,000 URLs or web pages were blocked in 2024 following government requests.

The amendments also introduce new rules for AI-generated content, defining for the first time audio and video created or altered to appear real, such as deepfakes, while excluding normal edits and genuine educational or design work.

Platforms must clearly label such content, add permanent markers to trace its source when possible, and cannot remove these labels afterward.

They must also use automated tools to detect illegal content, including deceptive material, fake documents, child-inappropriate content, explosives-related material, and identity impersonation.

Experts and digital rights groups have raised concerns about the feasibility and impact of these rules.

The Internet Freedom Foundation described the short timeline as turning platforms into “rapid-fire censors,” warning that it makes meaningful human review nearly impossible and may drive over-removal through automation.

Researcher Anushka Jain welcomed the labeling requirement for transparency but cautioned that the three-hour deadline could push companies toward full automation, risking excessive censorship.

Technology analyst Prasanto K. Roy called the new regime “perhaps the most extreme takedown system in any democracy,” noting compliance would be nearly impossible without extensive automation and minimal human oversight, with little time to assess the legality of each request.

Regarding AI labeling, Roy said the intention was positive but warned that reliable, tamper-proof labeling technologies are still under development.