While Washington argues about what to call AI, New Delhi is doing something more consequential: writing the rulebook and buying the chips to enforce it.
India's Union Minister Ashwini Vaishnaw announced on October 8 that the government will publish an AI regulation consultation paper within thirty days. Speaking at an event in New Delhi, he told his department to produce a document focused primarily on safety — covering AI safety, deepfakes, liability gaps, and skilling — as the foundation for a future techno-legal framework.
India's framing is deliberately lighter than Brussels. "We have always maintained that AI cannot be regulated by just having a law," Vaishnaw said. "We have to take a techno-legal approach," with the bulk of the responsibility falling on industry — especially on cybersecurity, safety, user harm, and deepfakes. An inter-ministerial AI Governance and Economic Group is already coordinating policy across ministries.
The consultation paper is expected to cover risk tiers for foundation models and high-impact applications, data localization expectations for sensitive sectors, and liability allocation between deployers and model providers. It is unclear whether the consultations will be public or restricted to selected stakeholders.
The second half of the announcement matters just as much. The government estimates demand for 10,000 GPUs, with a first tender covering 5,000, and is pursuing five-year supply agreements to give companies demand visibility. Instead of traditional tenders, the government will use an empanelment process: eligible AI models, solutions, and startups qualify once, then bid on departmental projects. India has already deployed more than 38,000 GPUs and wants roughly 100,000 by the end of 2026. Vaishnaw also said India is targeting indigenous GPUs for AI inference within three to four years.
The context is strategic autonomy. MeitY Secretary S Krishnan warned that privately owned foreign models pose risks — Indian data traveling abroad, foreign models learning from it — and officials are leaning toward open-source and open-weight models that can run inside India without sending data out.
India lacks a dedicated AI law; most governance falls under the Information Technology Act of 2000. A consultation paper does not change that overnight. But a 30-day deadline plus 10,000 GPUs is a statement of intent: New Delhi wants innovation-friendly rules, homegrown compute, and a seat at the global standards table — in that order.
