In a quiet but significant shift, the Trump administration has backed away from telling American AI companies to keep their newest models away from Britain’s safety testers. Here is what happened, and why it matters.
What exactly changed?
According to the Telegraph, reporting on Sunday, the administration has reversed a decision that asked US developers — companies such as OpenAI and Anthropic — not to share new AI models with Britain’s AI Security Institute before their public release. American labs can now give the British government early access for safety testing without running afoul of Washington.
What does Britain’s AI Security Institute actually do?
It is the UK government’s testing body for frontier AI, set up after the world’s first AI safety summit in 2023. Its job is to get its hands on the most advanced models before release and probe them for dangerous capabilities and misuse potential — the kind of evaluation that cannot be done from the outside once a model is already in the wild.
Why did Washington object in the first place?
The original request reflected the administration’s skepticism of international AI oversight. Trump has repeatedly dismissed calls for binding safety requirements on AI developers, and asking US companies to withhold pre-release models from a foreign government fit that posture: American technology, American rules.
So what moved?
The reversal lands just days after Trump’s newly launched AI task force said developers must report security incidents with their models — a requirement that, while far short of binding regulation, acknowledges that these systems need watching. The change signals a softer line on international safety cooperation, giving US developers more latitude to work with allies.
Has anyone confirmed this?
Not yet. Spokespersons for Trump, OpenAI and Anthropic did not immediately respond to requests for comment. For now the story rests on the Telegraph‘s reporting — worth noting, since neither government has put the new position on the record.
Why should anyone outside AI policy care?
Because the question of who gets to inspect the most powerful technology ever built — and when — is becoming one of the defining regulatory fights of the decade. Pre-release access is the difference between catching a dangerous capability in a lab and discovering it in the open market. Britain’s institute was designed to be the place where that happens. Whether Washington genuinely wants it to succeed, or simply decided the diplomatic cost of blocking it was too high, will become clearer the next time a frontier model is about to launch.
Related: Trump names intelligence chief to lead new ‘super intelligence’ AI task force · Anthropic bans users from being cruel to Claude
One Comment