I was halfway through a flat white in a Sydney cafe when my mate slid his phone across the table and said, "Bro, are you having issues with Fable?"
On Friday I'd been happily poking at Claude Fable 5, Anthropic's shiny new frontier model that had landed three days earlier. On Friday night it was glitchy, then suddenly it was not available. And the reason, it turned out, had nothing to do with my bill, my usage limits, or anything I'd done. It was because of who I am. A foreign national. A man with a New Zealand passport drinking coffee in Australia.
So let me tell you what actually happened, because it is one of the stranger things to land in our industry this year.
What actually happened?
On 12 June, at 5:21pm US Eastern time, Anthropic received an export control directive from the US government, citing national security authorities. The letter came from Commerce Secretary Howard Lutnick, written with help from the Commerce Department's Bureau of Industry and Security, and it ordered Anthropic to suspend all access to Fable 5 and Mythos 5 by any foreign national. Inside the United States or outside it. Customers, yes, but also Anthropic's own foreign-national employees (Anthropic).
Now, here is the practical problem with telling a company "only Americans may use this product." You cannot easily sort hundreds of millions of users into citizens and non-citizens overnight without making a hash of it. So Anthropic did the only thing it could to stay compliant. It switched both models off for everyone. The whole planet. Access to every other model, Opus 4.8 and the rest, stayed up (CNBC).
The trigger, as far as anyone can tell, was a jailbreak. The government believes someone worked out how to bypass Fable's safeguards. According to Axios, a rival company claimed it had jailbroken Mythos, which spooked the administration into action (Axios). Anthropic's version is rather more deflating. It says the government has so far offered only verbal evidence of a narrow technique that amounts to asking the model to read a codebase and fix the software flaws it finds. Anthropic reckons the same trick works on other publicly available models, including OpenAI's GPT-5.5, and that defenders use exactly this kind of thing every day to keep systems safe (Anthropic).
In other words, the most powerful AI on the market got recalled over something that, on Anthropic's account, your average security engineer does before lunch.
Why this is unusual?
Export controls are not new. We have spent years watching Washington and Beijing scrap over who gets which chips. What is new, and worth pausing on, is that this is the first time the export control tool, a thing designed for fighter-jet components and advanced semiconductors, has been pointed at a language model that was already sitting in the pockets of hundreds of millions of people (Fortune).
That is the bit that should make every technology leader sit up. A frontier model is no longer just a product. It is now, officially, a controlled good. And a controlled good can be turned off by a government you did not elect, for reasons you are not told, with roughly five hours and twenty-one minutes of notice.
Anthropic, to its credit, is not going quietly. It is complying with the order while flatly disagreeing with it, calling the whole thing a misunderstanding, and warning that if "we found one narrow jailbreak" became the bar for recalling a model, you would essentially halt every new model launch from every frontier lab forever (NBC News). They are probably right. That does not help me get my model back.
What to do next?
Here is the dirty secret of the whole "AI transformation" we have all been selling. The intelligence layer of your business now sits on infrastructure governed by foreign law. You can have the cleanest data, the best prompts, the slickest integration, and none of it matters if a letter lands on a Friday afternoon in another hemisphere.
This is not a reason to down tools and go back to spreadsheets. It is a reason to grow up about resilience. So, a few practical things worth doing while the dust settles.
- Map your single points of failure. Know exactly which workflows depend on one model from one provider. If the honest answer is "all of them," you have your first project.
- Enable Model Switching. Build a fallback into the architecture, abstract your model layer so you can swap providers without a rewrite.
- Read the data retention fine print. Anthropic now requires 30-day data retention on Mythos-class models as part of its safety strategy. That carries real cost and real compliance questions for regulated industries. Know what you signed.
- Ask the citizenship question before procurement, not after. "Can this be restricted by nationality" is now a legitimate line in a vendor risk assessment. A year ago it would have sounded paranoid.
- Keep a non-US option on the bench. You do not have to use it. You do have to know it exists, what it costs, and how long it takes to switch.
- Consider hosting your own LLM. It is possible and affordable to install your own LLM model on your own infrastructure. Yes it will need to keep being upgraded, but it will be your "sovereign" LLM.
The Bottom Line
We have spent two years being told these models are utilities, as reliable as electricity and water. This week one of them got cut off at the main for everyone outside one country, and the company that built it found out by letter. Utilities do not behave like that. Strategic assets in a geopolitical contest do.
None of this means the technology is any less remarkable, or that you should slow down. It means you should build like an adult who has read the news. Diversify your providers, abstract your model layer, and treat "available in your country today" as a feature that can be revoked.
“It is not your human right to have access to all models.”
As for me, I will be fine. There are other models, and the Fable will probably be back online by the time you read this. But I will not soon forget that the most advanced tool I have ever used decided, for one strange weekend, that I was a national security risk because of my passport.
References
Anthropic, Statement on the US government directive to suspend access to Fable 5 and Mythos 5, 12 June 2026 Bloomberg, Anthropic Says US Orders Halt to Foreign Access for Fable 5, Mythos 5 AI Models Axios, Scoop: Trump admin blocks foreign access to Anthropic's most powerful AI Fortune, Anthropic disables Fable and Mythos AI models after U.S. government bars it from giving foreigners access CNBC, Anthropic disables access to Fable 5 and Mythos 5 to comply with government directive NBC News, Anthropic suspends new AI models after government directive
