Britain lobbied Trump for an exemption from the Anthropic AI ban. The answer was no.

TL;DR

Britain lobbied the White House for an exemption from the Anthropic export ban and was told there was “zero chance.” The rejection exposes the UK’s dependence on American AI and strengthens the case for sovereign alternatives.

Sir Keir Starmer’s government spent the weekend lobbying the White House to restore British access to Anthropic’s most capable AI models. A source close to President Trump told The Telegraph there was “zero chance” of a UK carve-out.

The rejection lands as Starmer meets Trump at the G7 summit in Évian-les-Bains, where the ban is expected to come up directly. For thousands of British businesses, it is a sharp lesson in how quickly access to critical AI infrastructure can be switched off by a decision taken in another country.

What happened

On 12 June, the US Commerce Department ordered Anthropic to suspend foreign access to Fable 5 and Mythos 5, its most powerful models. Commerce Secretary Howard Lutnick sent CEO Dario Amodei a letter citing national security concerns over a reported jailbreak.

The 💜 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol' founder Boris, and some questionable AI art. It's free, every week, in your inbox. Sign up now!

The jailbreak was reportedly flagged by Amazon CEO Andy Jassy to senior administration officials. It involved asking the model to read a codebase and identify software flaws, a technique Anthropic says is available in rival models, including OpenAI’s GPT-5.5, and is used routinely by cybersecurity defenders.

Because Anthropic could not quickly build nationality-based access controls, it disabled both models globally, cutting off Americans and foreigners alike. “If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers,” the company said.

Why Britain is exposed

The UK has no frontier AI model of its own. Hospitals, financial firms, and government researchers that had integrated Fable 5 into their workflows found their tools abruptly disabled without prior warning.

Kanishka Narayan, the UK’s minister for AI, said the ban had direct implications for defence, noting that the most capable models are now used in drones, counter-drone systems, and cybersecurity. “The central question for our national security and defence is a question of our AI capability,” he said.

US Secretary of War Pete Hegseth struck a different tone. “Three months ago, the Department of War kicked Anthropic out of our building, forever,” he posted on X, adding that “every passing day proves why that was the right move.”

Industry pushback

More than 80 cybersecurity leaders at firms including Nvidia and Adobe signed an open letter urging the curbs be dropped. They argued the ban hampers efforts to find and patch software flaws and that Anthropic’s models represent only an incremental advance over rival systems, including China’s Kimi 2.7, which remains freely available.

On Monday, Anthropic met Commerce Department officials in Washington, joined by National Cyber Director Sean Cairncross, to negotiate a path forward. Export experts told Reuters the legal footing is shaky: AI models are reached through remote access, not physically exported, raising doubts over whether Commerce can lawfully impose the restriction.

The sovereign AI response

The UK government has already established a £500 million Sovereign AI Fund to back home-grown developers, and private capital is following with a separate £1 billion push for sovereign AI infrastructure. The Anthropic ban hands those efforts a powerful new justification.

At the G7, Canada’s Prime Minister Mark Carney compared the Anthropic shutdown to the 2008 financial crisis, warning of systemic “model risk” when critical infrastructure depends on a single provider. Insiders expect Trump will ultimately roll back the restriction globally rather than grant any country a bespoke exemption.

For every British firm that has wired a foreign AI model into its products and processes, the practical question is no longer theoretical. Resilience now means knowing which suppliers could be switched off overnight, and having a plan for the morning it happens.