Starmer seeks British carve-out from Trump’s Anthropic AI ban
Government in discussion with White House as it lobbies to be on approved countries list

Downing Street is hopeful that their sweeping intervention will prove temporaryCredit: Toby Shepheard/PA

James Titcomb
Technology Editor
Show biography
14 June 2026 2:05pm BST
Sir Keir Starmer is lobbying Trump’s administration to allow Britons to use Anthropic’s most advanced AI models after the White House banned foreigners from using the technology.

Government officials held discussions with the White House and Anthropic over the weekend after the US department of commerce demanded that the company’s Mythos 5 and Fable 5 models be cut off from non-Americans.

The shock ban has alarmed US allies after exposing how easily the crucial technology can be shut off.

However, Downing Street is hopeful that the sweeping intervention will prove temporary and be replaced by a more targeted regime.

Jade Leung, the Prime Minister’s AI adviser, is believed to have been coordinating discussions with the US and Anthropic about the possibility of including Britain on any future list of approved countries.

“They recognise there will be a more surgical approach to this [in future] and maybe a staged rollout,” said one source.

On Friday night, the US department of commerce served Anthropic with an export control directive banning all foreign nationals from accessing Fable 5, citing a “jailbreak” that could bypass its safety controls.

Advertisement

This would allow the system to be used for advanced hacking or, potentially, to develop bioweapons.

Anthropic, which disputes the severity of the issue, responded by shutting down access globally, saying this was the only way it could comply.

Restrictions were imposed in short order after Andy Jassy, Amazon’s chief executive, raised concerns with the White House that safety features within Anthropic’s models could be bypassed.

Last week, a security researcher known as Pliny the Liberator posted screenshots on X that appeared to show they had managed to get Fable 5 to share details on how to attack a Linux computer system and carry out stages of meth production.

The White House also acted against Anthropic in part because of fears that China had accessed the company’s systems, according to the US news website Semafor. Anthropic said this had not come up in discussions with the US government.

On Saturday night, Pete Hegseth, the US secretary of war, wrote on X: “Three months ago, the Department of War kicked Anthropic out of our building – forever. Every passing day proves why that was the right move.”

Advertisement

Mr Hegseth was referring to an earlier row with Anthropic about the extent to which its AI could be used by the armed forces, a dispute that led the US department of war to designate the company a supply chain risk.

There are hopes that the current situation will be resolved in the coming days, potentially leading to the order being lifted. But the use of export controls has raised concerns that Britain could be dependent on the whims of the White House to secure access to advanced AI.

While the Government’s AI Security Institute is seen as world-leading in safety testing, Britain does not have any AI giants of its own, control any part of the industry, or have sufficient infrastructure to train its own systems.

Kanishka Narayan, AI minister, wrote on X on Saturday: “As we debate the future of national security and technological sovereignty, access to AI capabilities is crucial.”

Advertisement

In an effort to kickstart the AI industry in Britain, Sir Keir last week announced that the Government would spend £400m on AI chips from British companies. The Government also plans to invest £500m in start-ups through a Sovereign AI fund.

However, there are doubts about how effective these efforts will prove to be, given the sums involved. Nvidia, whose chips are widely used to train and run AI systems, is worth almost $5tn (£3.7bn) and its main customers, including Amazon, Google and Microsoft, spend hundreds of billions of dollars between them on its chips.

The White House has previously delayed access to advanced AI outside the US. Until this month, non-American companies were forbidden from joining Project Glasswing, Anthropic’s cybersecurity project that gave businesses access to an earlier AI model.

The Trump administration, which had adopted a laissez-faire attitude towards AI, has become increasingly interventionist in recent weeks amid growing concerns about the capabilities of the latest systems.

Advertisement

A Government spokesman said: “We are aware of reports that access to Anthropic’s Fable 5 and Mythos 5 models has been paused worldwide, including in the US, following an export control directive from the US Government.
 
“We are in contact with both the US Government and Anthropic to understand the full picture and are monitoring the situation closely.”