“Too powerful,” boasted Anthropic. That’s what it called its own model, Claude Fable 5, days before it went public on June 9, 2026.
At 5.21pm, three days later, the AI company received a letter from the US government. The letter didn’t explain much. No detailed national security concern, no specific evidence. Just a directive, citing export control authority. It ordered Anthropic to cut off foreign access to Fable 5 immediately.
Anthropic posted a statement on its own website that same day. The company would need to “abruptly disable Fable 5 and Mythos 5 for all our customers”. By the next morning, both models, the one marketed as too powerful to release, and its sister model, Mythos 5, were gone. Everywhere. For every user.
Suggested Reading
Applying for a job with Steven Bartlett – very odd and very revealing
The US government had switched off a major American tech product, worldwide, overnight, without warning, and without ever fully explaining why. Uncharted territory.
The European Commission didn’t need explanations. Spokesman Thomas Regnier said the episode underlined “Europe’s need for technological sovereignty”. The commission unveiled measures that same month to slash the EU’s dependence on US and Asian tech.
But the story doesn’t end with AI hyperscalers. The same government mechanism that erased Anthropic’s flagship product overnight runs through Microsoft Office. That’s the software 345m businesses use to send email, hold meetings and keep the lights on. Could Trump flick the kill switch on Microsoft too?
Caisi is the Center for AI Standards and Innovation. It sits inside the US Commerce Department. It’s the body that pulled Fable 5 and Mythos 5.
It has a standing agreement with Microsoft, one of several it holds with major AI developers. It can test their models before and after public release. If Caisi doesn’t like what it finds, it can act. It produced the letter Anthropic received on June 12.
That’s not all. The AI Kill Switch Act, a bipartisan bill moving through Congress right now, would let the Secretary of Homeland Security order a shutdown of any AI system judged to pose “catastrophic harm”. Copilot, standard equipment across Microsoft’s business suite since July, is exactly that kind of system.
Trump’s Anthropic shutdown has put the notion of AI sovereignty squarely on the agenda for the world’s businesses and organisations. Tech writer Cory Doctorow believes the panic is overblown.
No government or serious business depends on ChatGPT or Claude for anything essential, and switching them off tomorrow would, in his words, do little damage. But he agrees a big risk sits in the everyday software running “administrative, telecoms and e-commerce” systems. If Trump ordered Microsoft to disable Office 365, that would have an immediate, measurable economic impact.
Would he actually do it? Absolute control, delivered instantly, with an audience watching? Evidence suggests Trump likes nothing better. Tariffs, sanctions, export controls, bombing raids. They are all blunt instruments he’s deployed with little consideration, warning, or justification.
American AI firms rely on foreign sales for their sky-high valuations. Every flicked kill switch makes the next foreign customer warier of buying in. Throw the switch once, and you don’t just hurt one company, you hand every nation a reason to shop elsewhere. It’s the kind of self-defeating action that would stop most leaders cold. But Trump doesn’t do cold calculation. He does the switch.
A year ago, the worst-case scenario a business planned for with a software vendor was a price hike, a data breach, a bankruptcy. Now there’s a new entry on that list: your vendor’s government might switch it off, for reasons that have nothing to do with you.
Suggested Reading
Why the Aussies are the only ones standing up to the tech giants
German enterprise software giant SAP provides the tech underpinning many of Britain’s biggest businesses. Its new agentic features run on Anthropic’s Claude by default. French AI Mistral sits there as an alternative, unused by most because they have yet to realise they’ve inherited a new, geopolitical dependency.
The US has always withheld arms, chips, and sensitive tech from other nations. Usually it was for reasons everyone could see coming – an adversary, a sanctions regime, a named conflict. What happened to Fable 5 and Mythos 5, nobody outside a small circle in Washington could have predicted.
This isn’t about export controls. It’s about chaos. Chaos means risk. And risk is something organisations and businesses can’t tolerate.
Trump has flipped the kill switch on one American tech product. Would you trust him not to do it again?
