The UK will keep getting a pre-release look at OpenAI's and Anthropic's most powerful models, not just a post-launch review.
Britain's AI Security Institute confirmed Monday it will continue testing frontier AI models from OpenAI and Anthropic before they ship, rather than only after. AI minister Kanishka Narayan said the UK secured that access by coordinating closely with Washington, Bloomberg reported. The announcement specifically reverses a US policy shift that had prioritized White House review over outside pre-release testing. For now, the UK's testing role with these two labs stays in place.
The headline framing is UK diplomacy winning a concession from the US, but the real story is about leverage. Pre-release testing only works if a government can see a model before the public does - once that window closes, oversight shifts to reacting after deployment rather than catching problems first. Narayan's comments suggest the UK sees that access as worth fighting to keep, even as the US apparently tried to centralize the process in the White House.
Whether this access becomes a durable standard or just a one-off diplomatic save is the question nobody's answering yet.