OpenAI and Microsoft established a joint safety review board to vet new AI models before public release, but GPT-4 testing in India occurred without clearance.

The reporting draws on more than 100 interviews and internal documents, including memos from OpenAI researchers Ilya Sutskever and Dario Amodei.

The timing aligns with Sam Altman’s 2026 call for an international AI governance body and India’s rapidly expanding role in AI, highlighted by about 180 million monthly ChatGPT users and a surge in generative AI app downloads in 2025.

India is described as a fast-growing AI market, with substantial monthly ChatGPT usage and strong growth in app downloads for generative AI tools.

OpenAI leadership and the safety board were not aware of the India test; a board member learned of it only during a six-hour meeting when an employee discreetly informed her.

The test was not disclosed to OpenAI leadership at the time, and board member Tasha McCauley was informed informally after a board meeting, with the CEO not raising it in that session.

Microsoft initially denied the India test to The New York Times, then confirmed that GPT-4 was tested on Bing in India after reporting by The New Yorker.

Microsoft secretly tested GPT-4 on Indian Bing users in 2022, making India the first real user-facing test market without participants' knowledge.

The New Yorker asserts India was the first real-world, unknowing test market for GPT-4 via Bing in 2022.

Sam Altman spoke in New Delhi on February 19, 2026, urging the creation of an international body to regulate artificial intelligence.

The deployment reportedly bypassed a joint OpenAI-Microsoft safety review, with OpenAI allegedly advising against the test, which Microsoft proceeded with without clearance.