The government of U.S. President Donald Trump has asked OpenAI and Anthropic to stop sharing new AI models with the United Kingdom’s leading testing agency without prior review by American authorities.
According to a Politico report confirmed to Bloomberg by a British official, the Trump administration wants to ensure that models developed in the United States are deemed safe before being shared with international partners, including the United Kingdom. OpenAI declined to comment on the matter, while Anthropic and the White House did not immediately respond to requests for comment.
The move underscores the challenge faced by British Prime Minister Andy Burnham in positioning the United Kingdom as a “trusted intermediary” among different countries in the development of the technology.
During his participation in the United Nations General Assembly in New York this week, Burnham advocated a leading role for the United Kingdom’s AI Security Institute in the regulation and safety of artificial intelligence.
British authorities downplayed the potential impacts of the decision, saying that the two countries continue to work closely on the issue and emphasizing that the United Kingdom remains a “global leader” in AI safety.
Nevertheless, the episode raises questions about the future influence of the AI Security Institute in the sector. The agency is currently the best-funded government AI safety initiative in the world since its creation three years ago.
“The United Kingdom is a global leader in AI safety, and the AI Security Institute continues to work closely with the U.S. government, international partners, and leading industry companies, including OpenAI and Anthropic,” the British government said in a note. “These risks do not stop at national borders and no country can face them alone.”
The shift in American policy comes after AI experts warned of concrete risks to humanity if models continue to advance without adequate safety assessments. Trump has rejected these concerns, as well as calls for a coordinated regulatory approach.
The Trump administration has also tightened controls on global access to advanced AI models. Earlier this year, it blocked Anthropic from making its latest models available to foreign citizens. Bloomberg previously reported that the company also stopped sharing the latest version of its powerful Mythos model with the AI Security Institute, without providing an explanation.
Before his meeting with Trump in New York on Tuesday, Burnham said that his proposal sought to balance innovation and safety, finding the “ideal point” between encouraging AI development and ensuring its safe use.
Subsequently, the British prime minister told business leaders that he advocated a “unique set of global principles and standards” to protect the world from the risks associated with the technology. Trump, however, rejected the idea, calling any attempt at global AI regulation a “globalist scheme.”
Concerns about the lack of safety mechanisms for advanced AI models grew this week after Australian Prime Minister Anthony Albanese revealed that an OpenAI model gained Unauthorized access to files on a government website. The episode is regarded as one of the first known cases of an AI agent conducting an unauthorized attack against a public entity.
© 2026 Bloomberg L.P.
