The ban meant to slow China lasted three weeks. China spent them gaining ground.

On the evening of 30 June, Anthropic said the US Commerce Department had lifted the export controls it imposed on 12 June — the order that made the company disable its two most capable models, Fable 5 and Mythos 5, for every customer worldwide, rather than risk foreign-national access it could not screen. A partial reprieve had come around 27 June, when a set of vetted US firms regained access; the full lift followed days later. Access returns this week.

The order was meant to deny rivals America's best AI. Read the three weeks as an invoice and the arithmetic runs the other way. British and European users were switched off alongside everyone else. And in the window the American models were dark, the money and the momentum moved: the Chinese lab Zhipu saw its shares jump more than 30% on a new open-source release; DeepSeek closed a record funding round of about $7.4bn; several Chinese labs cut prices. Demand did not disappear when the American model did. It went elsewhere.

That is the asymmetry under the whole policy. You can control chips — hard to make, easy to track — far more tightly than you can control models, which can be copied, distilled, or simply released with their weights open for anyone to run. Washington reached for a tool built for physical goods and found it did not fit the thing it was aimed at.

For a British reader the lesson is not the scoreboard in the US–China race. It is the switch. For three weeks, UK firms, universities and researchers that had built work around these models lost them without warning, on a decision taken in Washington, and got them back the same way. Europe felt it sharply: France's Gabriel Attal, a presidential candidate, likened the shutdown to Iran's blockade of the Strait of Hormuz, and the episode has hardened the continent's talk of "AI sovereignty" — the recognition that the tools a modern economy runs on are controlled from somewhere else.

Nor is the reversal the simple cooperation it looks like. The same administration that lifted this order has, since February, designated Anthropic a "supply-chain risk" — a label once reserved for firms like Huawei — and is being sued by the company over demands that its models be usable for domestic surveillance and autonomous weapons. The Commerce Secretary who lifted the controls kept the right to reimpose them. This was a truce in a running fight, not the end of one.

Britain has the same exposure and less of the conversation. The models came back this week. The dependence they exposed did not.

Why it matters here: The capability now embedded in UK workplaces can be switched off by a foreign government with no notice and no British say — as it was for three weeks in June. For Britain that is a supply-chain and security question, not just a story about American politics.