US drops objection to Britain testing American AI models before release
The Trump administration has reportedly reversed a request that US developers withhold new AI models from Britain's AI Security Institute ahead of launch.

American artificial intelligence developers are once again free to give the British government early access to their models for safety testing without crossing the Trump administration, according to a report published by The Telegraph on Sunday, October 11, 2026. The administration has reportedly withdrawn an earlier decision that discouraged such cooperation.
Under that earlier stance, Washington had asked companies including OpenAI and Anthropic not to hand new models to Britain's AI Security Institute before they were publicly released. With the request now rescinded, US firms can work with British evaluators on pre-launch testing without fear of clashing with the White House.
The reversal has not been confirmed in an official statement. Spokespeople for Trump, OpenAI and Anthropic did not immediately respond to requests for comment.
What the British institute does
The AI Security Institute was established after Britain hosted the world's first AI safety summit in 2023. It examines the most advanced AI systems to gauge their capabilities and the risk that they could be misused. Access before release matters because it lets testers flag dangerous behaviour before a model reaches the public.
The move is being read as a sign that the Trump administration is softening its position on international AI safety work. It gives American developers more room to collaborate with London at a time when governments are still struggling to decide how this fast-advancing technology should be overseen.
A shifting position in Washington
Trump has repeatedly rejected calls for binding safety obligations on AI companies, but that position now appears to be evolving. On Friday, October 9, 2026, his newly created AI task force said developers are required to report security incidents involving their models.
Taken together, the two developments suggest a more accommodating approach both to domestic oversight and to cooperation with allies. Whether they add up to a lasting change in US policy will depend on what the administration does next.
Comments
No comments yet. Start the discussion or ask a question.
Related
Microsoft President Brad Smith urged urgent global AI safety rules, proposing emergency brakes, mandatory incident reporting and independent pre-release audits.
