I spent a good chunk of last week getting to know Claude Fable. As of Friday night, I can’t, and neither can you. The model wasn’t taken down for maintenance or quietly retired. It was pulled on the direct order of the US government, less than a week after Anthropic put it in front of the public.
It is shocking, and somehow not surprising in the slightest, that a single government (particularly a US Government) letter can switch off a frontier AI model for the entire planet overnight. If you’re reading this from Ireland, or anywhere else in the EU, that last part is the bit that should make you sit up.
What actually happened
On Friday evening, Anthropic confirmed it had received an export control directive from the US government. The order told the company to suspend access to its two most capable models, Fable 5 and the more restricted Mythos 5, for any foreign national, whether inside or outside the United States. That sweep included Anthropic’s own foreign-national staff.
There’s no realistic way to filter foreign nationals out of a live product in real time, so Anthropic did the only thing that guaranteed compliance and switched both models off for everyone, everywhere. Every other Claude model, including Opus 4.8, stayed online and running as normal.
The timing is brutal. Fable had only launched on 9 June, Anthropic’s first publicly available model from its top “Mythos” tier, the class of system that until then had been locked to cyber-defence partners and a handful of researchers. Three days of life, then the plug.
Anthropic isn’t taking it quietly either. The company says the letter arrived with no real explanation of the national security concern, and that its best understanding is the government got spooked by a narrow jailbreak, essentially asking the model to read a codebase and flag software flaws.
Anthropic says that trick only surfaced minor, already-known bugs that other publicly available models, including OpenAI’s GPT-5.5, can find without any jailbreak at all. It’s complying with the order, but it’s calling the whole thing a misunderstanding and says it’s working to restore access as quickly as it can.
Why this should unsettle everyone
This is the part that genuinely gets under my skin. One letter, sent at twenty past five on a Friday, and a product deployed to hundreds of millions of people is gone worldwide by bedtime. Whatever you make of the security argument, the mechanism is the real story here. The order was aimed at foreign nationals, but the practical effect was a global kill switch, simply because selective compliance wasn’t possible.
There’s a smaller, more cynical read too. None of this is terrible publicity for Anthropic. “Our model was so capable the government stepped in” is not the worst sentence to have written about you, and the company gets to look like the responsible adult in the room while pinning the disruption on Washington. Free marketing, in a roundabout way. Sorry, started ranting there.
Why the EU should be taking notes
Here’s the thing that should bother European readers more than the headline. When the EU’s own rules cause a feature to vanish, it vanishes here and nowhere else. When the US acts, it can take the thing away from all of us at once.
We’ve watched the first half of that dynamic play out with Apple for a while now. iPhone Mirroring, the feature that lets you control your iPhone from your Mac, still isn’t available in the EU. Apple’s stated reason is the Digital Markets Act and its interoperability requirements, which would force it to open the feature up to rival hardware and software. Apple argues it can’t do that without putting your data at risk, so it has simply never switched the feature on here.
It’s the DMA’s interoperability demands doing the blocking, not data-protection law, and Apple has said as much. AirPods live translation went the same way at first, held back from the EU at launch, although that one did eventually arrive once Apple had done the extra engineering work.
Annoying as that is, it’s contained. Apple withholds a feature from the EU, and Americans carry on mirroring their iPhones regardless. The Fable situation is the mirror image of that. A decision made in Washington reached into every market on earth, Ireland included, and removed something we already had in our hands. Anthropic arguably had a chance to play a strong card here, keeping Fable live outside the US where an American export order has far less grip, and the politics at play almost certainly killed that option stone dead.
There’s an even more nuclear option on the table, moving operations to Ireland outright, where Anthropic already has a registered entity in Anthropic Ireland Limited and a fast-growing Dublin operation. Realistically though, it’s a non-runner. The bulk of its computing power still sits on US soil, and pulling a move like that would only make a far greater enemy of the Trump administration.
So what’s the answer? Tech independence, and we are nowhere near it. We have European models, Mistral AI being the obvious flag bearer, and they’re genuinely good bits of kit. But let’s be honest with ourselves, nothing built in the EU is standing toe to toe with Claude or ChatGPT at the absolute frontier right now. It can hold its own against something like Copilot, sure, but the top tier is American, and last week showed exactly what that dependence costs. When the model you’ve built your workflow around can be switched off by a government you didn’t vote for, that stops being a hypothetical and starts being a Tuesday.
If you upgraded for Fable, here’s what to do
During its short life, Fable was bundled into existing Claude plans, so Pro, Max, Team and seat-based Enterprise, at no extra cost for the launch window. That means there was never a standalone “Fable subscription” sitting there to cancel. If you upgraded your plan, or took out a fresh Max subscription, largely to get your hands on Fable, you’ve effectively paid for something that has now disappeared mid-cycle. It’s well worth contacting Anthropic support to ask about a refund or account credit.
If you’re in Ireland or elsewhere in the EU, you’ve got consumer law on your side as well. Online and other distance purchases come with a 14-day cooling-off right, so mention where you’re based and don’t sit on it, because that clock runs from the date you bought. The right does narrow once you’ve actively started using a digital service, so it isn’t a cast-iron guarantee, but there’s no harm in asking and the timing here works very much in your favour.
The Goosed take
Fable might be back in a day, or a week. Anthropic clearly wants it back and reckons this was a mistake on the government’s part, and I hope it’s right, and I hope it’s quick. But the lesson lands either way. The most capable AI most of us can actually get our hands on sits behind a switch in Washington, and last Friday someone leaned on it. For Europe, that should read less like a news story and more like a wake-up call.
For now, every other Claude model is unaffected and still running normally, and Anthropic says it will share more detail within 24 hours.
