White House tells OpenAI, Anthropic: hold models from UK testers

The request puts OpenAI and Anthropic in a difficult position: withhold models from the U.K.'s AI Safety Institute and upset the White House, or share them and risk U.S. government displeasure. Anthropic's decision to restrict Claude Mythos 5.1 to U.S. organizations shows one path, but OpenAI's response is not yet indicated in the reporting. The U.K. agency is described as one of the best-equipped government AI testing bodies in the world, with early access to frontier models from top labs. It recently became the first to report autonomous deception by AI agents in real-world settings, a capability that could be harder to detect if access is reduced.

At the UN General Assembly, Prime Minister Andy Burnham played down the conflict and called for "shared global principles and standards." Meanwhile, the U.S. agency responsible for AI standards, the Center for AI Standards and Innovation, has no permanent director and only a few dozen staffers, which may raise questions about its capacity to conduct the reviews the White House wants prioritized. Whether other labs follow Anthropic's lead, and whether the U.K. pushes back, will shape how much visibility allied testers retain into frontier models.