India tightens AI rules

India is moving from “please be careful with artificial intelligence” to “label it, store it, and take it down fast.” A Ministry of Electronics and Information Technology draft released on March 30, 2026 would make ministry advisories and operating instructions part of the legal due-diligence checklist for platforms, with public comments due by April 14. (meity.gov.in) That matters because India already tightened its rules on February 10, 2026. The notified amendment that took effect on February 20 requires prominent labels for photorealistic artificial-intelligence content and cuts some takedown windows from 24 to 36 hours down to 2 to 3 hours. (thehindu.com) The government’s term for this material is “synthetically generated information.” In the October 22, 2025 explanatory note, the ministry defined it as content artificially created or altered by a computer in a way that appears authentic, with deepfakes, fabricated political clips, and fraud listed as examples. (meity.gov.in) The February rules were mostly about the content itself. Platforms must seek disclosure from users when content is artificial-intelligence generated, and if that disclosure is missing they may have to label it themselves or remove it when it is a non-consensual deepfake. (thehindu.com) The March draft goes wider than deepfakes. It adds a new Rule 3(4) saying an intermediary must comply with any ministry clarification, advisory, order, direction, standard operating procedure, code of practice, or guideline issued in writing about how Part II of the rules should work. (meity.gov.in) That is where safe harbour comes in. Safe harbour is the legal shield in Section 79 of India’s Information Technology Act that keeps a platform from being treated like the publisher of every post, and MediaNama notes the draft would tie that shield more tightly to compliance with ministry-issued instructions. (medianama.com) The other big shift is about who counts as “news.” The draft would make Part III apply not just to formal news publishers but also to intermediaries and “news and current affairs content hosted by non-publisher users,” which means an ordinary account posting commentary or reporting could get pulled into a framework built for digital news outlets. (meity.gov.in) Part III is not a light-touch section. Under the current structure, it connects to the Ministry of Information and Broadcasting’s Code of Ethics and to Rules 14, 15, and 16, which cover committee review, blocking directions for specific content, and emergency blocking. (medianama.com) The draft also changes how cases can start. The Inter-Departmental Committee would no longer look only at complaints from the public, because the March text says it can hear matters referred by the ministry itself. (meity.gov.in) One more clause looks small but is not. The draft says data-retention duties under the platform rules are “without prejudice” to preservation duties under any other law, which means a company cannot point to the usual 180-day limit if some other legal demand says keep the data longer. (meity.gov.in) Digital-rights groups are reading this as a power shift from published rules to day-to-day executive instructions. IFEX said the proposal could increase opacity in platform governance and chill free expression, while The Economic Times reported similar concerns that safe-harbour protection could hinge on following a widening set of ministry directives. (ifex.org) (economictimes.indiatimes.com) So the story is not just that India wants labels on artificial-intelligence images. It is that after the February 2026 deepfake rules, the March 2026 draft would give the government a tighter grip on how platforms interpret speech, how fast they act, what they must store, and which user posts can be treated like regulated news content. (thehindu.com) (meity.gov.in)