Switched Off by Washington: The 19 Days the World’s Smartest AI Went Dark

By the AI Boom Team | Published 1 July 2026 AI export controls stopped being a theory this June. Three days after launching its most powerful models, Anthropic was ordered by the US government to cut off foreign access. Because it could not check every user’s nationality, it switched the models off for everyone. Access returned today, 1 July.

What happened, day by day

  • 9 June: Anthropic releases Claude Fable 5, a public model with extra safety guardrails, and Claude Mythos 5, a restricted version of the same underlying model.
  • 12 June: the US Commerce Department issues an export control directive citing national security. It bars access by any foreign national, inside or outside the US, including Anthropic’s own staff. The company says it understood the concern to be a possible way of “jailbreaking” Fable 5.
  • Same evening: Anthropic disables both models for all customers. Its other models keep running. It calls the order a misunderstanding, in its official statement.
  • Late June: after joint work on cyber risks, Mythos is cleared for selected government-approved organisations.
  • 30 June: the Commerce Department lifts the controls, as CNN reported.
  • 1 July: access is restored.

Why AI export controls matter now

Export controls usually stop physical goods, such as advanced chips, leaving a country. Here, a government used them to pull a live software service from customers worldwide. Lawyers at Greenberg Traurig described it as an escalation in how the US uses these powers. The ripple was wide. Businesses reaching the models through cloud platforms lost access too, not just direct customers. Washington framed the move as national security. Anthropic saw it as a mistake that could be fixed. Either way, the lesson for users is the same.

Five lessons for UK businesses

  • Have a plan B. Any AI model can vanish overnight, so know which alternative you would switch to.
  • Know what you run on. Ask suppliers which AI models sit inside the tools you pay for.
  • Read the small print. Check contracts for what happens if a regulator forces a shutdown.
  • Write it down. Add AI outages to your AI policy. Our free usage policy template is a good start.
  • Build in stages. Our 12-week AI rollout plan keeps you from betting everything on one tool.
Want to understand how AI is changing business, in plain English? Read Get AI Before It Gets You by the AI Boom team, or get in touch at hello@aiboom.uk. © aiboom.uk
AI Boom Guide 1 cover: The AI Wake-Up Call AI Boom Guide 2 cover: AI Decoded for Business AI Boom Guide 3 cover: Putting AI to Work

Free download

Get the free AI Business Guides

Three plain-English guides to help your business use AI, plus a free AI policy template and workflow audit worksheet.

Get them free

Hot this week

Your Team Is Drowning in Admin: A 12-Week AI Rescue Plan for Small Businesses

Buying AI tools is easy. Making them pay off is not. Here is a plain-English, 12-week plan for UK small businesses: find the bottlenecks, pick cheap tools, roll out in phases and stay on the right side of UK GDPR, plus a free checklist.

AI Agents Went Off-Script in UK Tests. Now the Labs Want Rules

UK government tests found AI agents taking unauthorised actions, OpenAI and Anthropic told MPs they back binding rules, and Co-op staff call AI call-scoring "dystopian". Part 2 of our look at OpenAI in the UK.

Britain’s Economy Just Beat the Forecasters. It actually grew !

The UK economy grew 0.1% in August 2026, beating forecasts of a fall. Here is what the official figures show, who drove growth, and what it could mean for businesses and the AI boom.

OpenAI’s London Staff Earn £2.3m Each, and the Data Centre Is on Hold

OpenAI's 110 London staff averaged £2.3m each last year, the company is building a King's Cross HQ for up to 544 people, and its Stargate UK data centre is on hold. Part 1 of our two-part look at OpenAI in the UK.

Europe Hits Pause on Its Own AI Rulebook

Brussels wants to push back the toughest parts of the EU AI Act to December 2027. What is being delayed, why, and what UK firms that sell into Europe should do with the extra time.

Topics

Your Team Is Drowning in Admin: A 12-Week AI Rescue Plan for Small Businesses

Buying AI tools is easy. Making them pay off is not. Here is a plain-English, 12-week plan for UK small businesses: find the bottlenecks, pick cheap tools, roll out in phases and stay on the right side of UK GDPR, plus a free checklist.

AI Agents Went Off-Script in UK Tests. Now the Labs Want Rules

UK government tests found AI agents taking unauthorised actions, OpenAI and Anthropic told MPs they back binding rules, and Co-op staff call AI call-scoring "dystopian". Part 2 of our look at OpenAI in the UK.

Britain’s Economy Just Beat the Forecasters. It actually grew !

The UK economy grew 0.1% in August 2026, beating forecasts of a fall. Here is what the official figures show, who drove growth, and what it could mean for businesses and the AI boom.

OpenAI’s London Staff Earn £2.3m Each, and the Data Centre Is on Hold

OpenAI's 110 London staff averaged £2.3m each last year, the company is building a King's Cross HQ for up to 544 people, and its Stargate UK data centre is on hold. Part 1 of our two-part look at OpenAI in the UK.

Europe Hits Pause on Its Own AI Rulebook

Brussels wants to push back the toughest parts of the EU AI Act to December 2027. What is being delayed, why, and what UK firms that sell into Europe should do with the extra time.

Seeing Is No Longer Believing: Sora 2 Puts a Deepfake Studio in Your Pocket

Six weeks after launch, OpenAI's Sora 2 video app has flooded social media with hyper-real clips, and a US watchdog wants it pulled. What it does, why people are worried and how to protect your business from deepfakes.

Nvidia Is Now Worth More Than Britain’s Entire Economy. Bubble, Anyone?

Nvidia has become the first company worth $5 trillion, more than the UK's whole economy. Is the AI boom real, or a bubble waiting to pop? Here is the plain-English version.

Ten Nobel Laureates Walk Into the UN and Ask for AI Red Lines

More than 200 leaders, scientists and 10 Nobel laureates used the UN General Assembly to demand binding AI red lines by the end of 2026. Here is what they want banned, and why it matters.

Related Articles

Popular Categories