MeitY unveils tougher rules for AI content
If the complaint involves non-consensual intimate images, nudity or impersonation, the take-down time is 2 hours
The Union Ministry of Electronics and Information Technology (MeitY) has suggested changes to the information technology rules for faster online content takedowns. If these changes take effect, social media platforms and other intermediaries will have to remove unlawful, AI-generated content within three hours of getting a valid Government or court order. The current deadline is 36-hours.
The updated rules also require clear labels on AI-generated content and faster responses to complaints. These steps aim to tackle problems caused by deepfakes, AI-driven misinformation, and other types of synthetic media.
In February 2026, MeitY announced changes to the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021. The new rules set up a formal process for handling Synthetically-Generated Information (SGI), including audio, visual, or audio-visual material made or changed by AI or algorithms to look real. The proposed changes shorten the time online platforms have to respond.
If the complaint involves non-consensual intimate images, nudity or impersonation, the deadline is two hours. Platforms have to handle general complaints within seven days, down from 15, and urgent cases within 36 hours. They also need to label approved SGI so users can clearly spot synthetic content. When possible, permanent metadata or unique identifiers should be added to help trace content.
Routine good-faith edits, accessibility enhancements, and certain educational or conceptual uses are excluded from the definition of SGI.
Non-compliance with these regulations may result in the loss of safe-harbour protection under Section 79 of the IT Act.
Officials say the old 36-hour response time was not enough once content went viral. The new labelling rules are meant to help users tell real content from synthetic material.
The rules also make it clear that SGI counts as ‘information’ under current laws, covering an area that was not regulated before.
A 24-hour deadline for all unlawful content, even beyond high-risk material, leaves little time for thorough human review. Because of this, platforms may rely more on automated systems, which could lead to the removal of lawful speech, parody, satire, or political commentary. It is hard to tell the difference between deepfakes and protected expression in such a short time. Costs for following the rules are expected to go up, and keeping up with moderation, detection tools, and persistent identifiers will be tough, especially for smaller platforms.
Observers note that the final rules replaced the draft’s 10 per cent visual-area labelling rule with a more flexible ‘prominent’ standard, allowing for technical and creative differences. Even with this change, the shorter takedown deadlines are still hard to meet. The rules now require reasoned notices instead of simple orders, but the faster timeline gives platforms or users little chance to challenge orders before content is removed. Free-speech advocates worry that platforms might remove content too quickly to keep safe-harbour protection. Consistent enforcement by officials and courts is needed to avoid unfair results. These changes are India’s most detailed rules for generative AI and online platforms so far.
