India: the government puts ministries on pause on OpenAI and Anthropic

In plain terms

The Indian Ministry of Electronics and IT (MeitY) has asked its ministries to suspend the deployment of OpenAI (ChatGPT) and Anthropic (Claude) models in cybersecurity-related uses, pending an audit. Officially temporary and targeted—in practice, it's a strong political signal.

The context

Since 2024, India has built its AI policy around a simple idea: sovereignty starts with data. IndiaAI Mission (₹10,372 crore, March 2024) subsidizes local compute, public corpora, and the BharatGen foundation model. Bhashini, the government's open-source initiative, has positioned itself as an offline testable stack. Sarvam AI has raised a significant round with an explicit mandate: to build an LLM operated in India, for Indian languages, on domestic infrastructure. Krutrim (Ola) is playing the same card.

The data

According to Medianama (July 13, 2026), the MeitY note targets the use of OpenAI and Anthropic models for cybersecurity use cases in government agencies and systems, as part of a tightening of controls on external AI tools, sensitive data, and public systems. No expanded scope to the rest of the workflows is confirmed. No resumption date is published.

The analysis

Two interpretations coexist. The first, official: a standard security audit, in reaction to a general tightening of controls on external AI tools in public systems. The second, structural: it's the expected lever to force the adoption of a sovereign stack—the logical follow-up to Bhashini. The timing—in the wake of Zhipu's "Touch High" plan and Sarvam's funding—is not trivial. Delhi is watching Beijing and signaling to Washington that its government AI stack will not be American by default.

So what

For a CIO of a group operating in India: map calls to US APIs on the public sector side, prioritize sensitive workflows, and anticipate an "India-only" endpoint. For an agent founder: the "API sovereign" stack (Sarvam, Krutrim, Llama on-prem) becomes a contractual selection criterion in the Indian public market—less optional.