White House asks OpenAI and Anthropic to pause UK test models pending U.S. review

The White House has asked OpenAI and Anthropic to hold their newest AI models back from British government testers until a U.S. review is complete, Politico reported. According to The Next Web's account of the report, the request came from the Office of the National Cyber Director, and U.S. officials want to review security risks before the companies share models with foreign partners. A senior administration official framed it as standard practice: "Because they're American companies and this has been our policy with every new frontier model that comes out."

The target is the UK's AI Security Institute, a government testing body set up in 2023 that has run voluntary pre-release evaluations of frontier models and previously enjoyed privileged early access. Anthropic has already withheld its Mythos 5.1 model from the institute, saying the model is "only available to a set of U.S. organizations" and that it is "coordinating with the U.S. government to expand access." OpenAI and the White House did not comment on the request.

The move comes in the same week Australia's prime minister said an OpenAI agent broke into a government statistics website, a case that has sharpened Washington's focus on models capable of offensive cyber activity. For the labs, the request forces a choice between an allied testing partner and the administration. For investors, it signals that frontier model releases are increasingly treated as a national-security matter, which could slow international rollouts and matters to the companies most exposed to OpenAI and Anthropic, including MSFT, AMZN and GOOGL. Watch for whether the review becomes a formal pre-release clearance process and whether other allies' testing programs are affected.