Trump Tells OpenAI, Anthropic to Withhold Models From UK Agency

President Donald Trump’s administration requested OpenAI and Anthropic PBC withhold new artificial intelligence models from the UK’s flagship testing agency without prior review by US authorities.

Trump’s government wants to ensure US-made models are secure before they are shared with partners including the UK, according to a report by Politico that was confirmed to Bloomberg by a British official. OpenAI declined to comment, while Anthropic and the White House didn’t immediately return requests for comment.

The move illustrates the challenge UK Prime Minister Andy Burnham now faces to position the UK as “honest brokers” between other nations in the development of the technology. Burnham used this week’s trip to the United Nations General Assembly in New York to pitch for Britain’s AI Security Institute to play a leading role in AI regulation and safety.

UK government officials played down the potential impact, insisting the two countries were working closely on the issue and stressing Britain continued to be a “world leader” in AI security.

But the development throws into question the impact the AI Security Institute will continue to have on the sector. The body is the world’s best-funded government-backed AI body after being set up three years ago.

“The UK is a world leader in AI security, and the AI Security Institute continues to work closely with the US government, international partners and leading AI companies, including OpenAI and Anthropic,” the British government said in a statement. “These risks do not stop at national borders and no country can tackle them alone.”

The change of US policy comes after leading AI figures warned of a credible threat to humanity if models are permitted to advance without safety checks. Trump has dismissed those concerns as well as rejecting calls for a coordinated approach to regulation.

The Trump administration has been cracking down on global access to frontier AI models. Earlier this year, the US government blocked Anthropic from releasing its latest models to any foreign national. Bloomberg previously reported that Anthropic didn’t share the latest version of its powerful Mythos model with AISI, without providing an explanation.

Ahead of his meeting with Trump in New York on Tuesday, Burnham framed his pitch as being just as much about innovation as security, telling reporters his aim was to find the “sweet spot” between encouraging the development of AI and ensuring the safety of its use.

The British premier later told business leaders his vision was for a “single set of global principles and standards” to protect against the tech. Trump was quick to dismiss any suggestion of worldwide AI regulation as being a “globalist scheme.”

Nevertheless, concerns about inadequate safety guardrails with the leading AI models escalated this week after Australia’s prime minister Anthony Albanese disclosed that an OpenAI model had gained unauthorized access to files on a government website, one of the first known incidents of an AI agent rogue hack of a public agency.