White House Reverses the 17-Day UK AISI Freeze: Frontier-Model Access Restored — and London Formally Announces It, With a Warning About Permanence

White House Reverses the 17-Day UK AISI Freeze: Frontier-Model Access Restored — and London Formally Announces It, With a Warning About Permanence

Update (Oct 12): the reversal is now formally announced by the UK government, with new detail on how it happened and what it means. At a speech at Imperial College London on Monday, AI Minister Kanishka Narayan announced that the UK's AI Security Institute "will resume receiving access to the most powerful AI models from OpenAI and Anthropic before their public release, reversing a U.S. policy that had cut off such access" (Quartz, citing the gov.uk announcement).

The original story (reported Oct 11): the Trump administration's Sept 24 ONCD national-security order had frozen Britain's access to unreleased frontier models for 17 days — OpenAI and Anthropic were told to withhold models from AISI until US authorities cleared them¹. Per the Telegraph (via LBC/Bloomberg), the freeze led Anthropic to stop British officials from reviewing its Claude Mythos 5.1 model, which stayed US-only, with no explanation given.

How the reversal happened: Narayan credited lobbying and recent minister-to-executive conversations "facilitated in part by a summit with King Charles III" (Bloomberg, via Quartz). His quotes cut both ways:

"I'm very confident that, as a consequence of the conversations we've had, we will be able to test the latest generation of models."

"When you're reliant on someone else's technology, you have to be clear-eyed about the permanence of any outcome.²"

The rest of the speech shows what London is doing with the restored access: two government pilots using frontier AI "identified more than 1,000 potential vulnerabilities in government systems," and Narayan said he would work with other ministries to find weaknesses before attackers do. His governing line: "If you cannot adequately control the risks, do not release the system."

Context: the dispute sits inside a larger UK–US governance rift — PM Andy Burnham casts the UK as an "honest broker" on global AI standards with AI a focus of its G20 presidency, while Trump says he will resist efforts to constrain "super intelligence."

Still open: no written framework has been confirmed — the arrangement remains effectively an informal understanding that a future US directive can suspend, exactly as the 17-day freeze proved. Whether the national-security review's findings ever surface, and whether Anthropic now expands Mythos 5.1 beyond the US-only launch distribution, remain unanswered.