What happened
Reports dated 28 October 2026 (citing officials, NDTV and Mint) say India's Ministry of Electronics and Information Technology (MeitY) is preparing a consultation paper for a standalone AI law. Under consideration is a separate safety framework for 'frontier models' — pre-release evaluations, controlled-environment testing, and sandbox testing proportionate to capability, applied before a model believed capable of serious harm is released. Officials also flagged the need to define an 'AI intermediary' separately from social-media intermediaries, which has implications for safe harbour and liability for AI-generated output. IT Minister Ashwini Vaishnaw has said the consultation paper will be published in the coming weeks (possibly November 2026) and will seek input from industry, civil society and citizens.
Why it matters
This signals India's pivot from voluntary, technology-neutral guidance towards legislation that could mandate pre-deployment safety testing for the most capable models and clarify intermediary/safe-harbour liability — affecting any frontier or enterprise AI provider operating in one of the world's largest AI markets.
Action needed
Frontier and enterprise AI providers operating in India should track the forthcoming consultation paper (expected November 2026) and prepare submissions on the frontier-model definition, sandbox/pre-release evaluation requirements, and the AI-intermediary safe-harbour question.