Britain Loses First Look at America's Newest AI Models

The White House has asked OpenAI and Anthropic not to give their newest models to the UK's AI Security Institute until US agencies have tested them, Politico reported on 24 September 2026. The request came from the Office of the National Cyber Director. Politico attributed it to a person familiar with the matter and a senior US administration official, both granted anonymity to describe it, and said the White House did not respond to a request for comment.

The model that did not cross the Atlantic

Anthropic has gone along with the request so far. Its Claude Mythos 5.1 was, in the company's own announcement, "only available to a set of U.S. organizations", and Anthropic said it was "coordinating with the U.S. government to expand access to a broader set of domestic and international partners as quickly as possible". Anthropic declined to comment to Politico.

The gap is documented in a primary source. In a letter dated 15 September 2026 to Liam Byrne, who chairs the House of Commons Business and Trade Committee, AISI director Henry de Zoete wrote that "Anthropic made clear at the time of the release of Mythos 5.1 that no organisations outside of the US had access to the model". He added that the institute remains "in daily conversation with industry partners, including Anthropic".

De Zoete described a narrowing rather than a closure. "We maintain strong relationships with all frontier AI developers and continue to have pre-release access to some of the world's most capable models," he wrote, noting that AISI had tested OpenAI's GPT-6 Astra before its public release. Parallax Nexus covered that release, and the decision to route exploit-class capability only to vetted defenders, earlier this month.

Washington's stated reason

The same official said the administration wants the United States to review the models and make sure US systems are secure before they are shared with partners. A British official confirmed the substance of Politico's report to Bloomberg the following day. OpenAI declined to comment to Bloomberg, and the White House replied to neither outlet, so nobody who issued the request has explained it on the record.

Two speeches in one week

The timing is awkward for London. Prime Minister Andy Burnham told the UN General Assembly that the institute was "working hand in glove with the US and the leading labs on AI safety" and called for a "single set of global principles and standards". Foreign Secretary Ed Miliband told the UN Security Council on 23 September 2026 that governments need to ensure frontier models are "rigorously tested", adding that the leading companies had committed to provide that visibility.

A UK government spokesperson said that "these risks do not stop at national borders and no country can tackle them alone".

The agency that would go first instead

The body Washington wants at the front of the queue is the Commerce Department's Center for AI Standards and Innovation. Politico reports that it is operating without a permanent director and with only a few dozen technical employees, while being asked to evaluate a growing stream of frontier models. De Zoete's letter describes AISI as "the largest and best resourced AI security institute in the world". On those two descriptions, the practical effect of the request is to move first-look testing from the better-staffed agency to the thinner one.

The honest reading is that pre-release access was always a favour rather than an entitlement, granted voluntarily and withdrawable by the companies or by their government, and that this week settled who decides. Against that, AISI did test the most capable OpenAI model released this month, which points to a narrowed pipeline rather than a severed one. What is not known is whether OpenAI has agreed at all, whether this is written policy or an informal ask, and how long a US review takes.

In the same letter, de Zoete offered the committee a visit to AISI on 12, 13 or 15 October. That gives MPs a date on which to ask, on the record, who decides what Britain is allowed to test.