Anthropic's Claude 'Fable 5' Access Ban: Full Story Finally Told — And Refund Deadline Is Today

Here's a story that shows just how much the government's hand can reach into the AI tools you use every day — and how fast things can disappear.

On June 9, 2026, Anthropic launched what it called its most powerful AI model ever made publicly available: Claude Fable 5. Think of it as the consumer-facing version of an even more powerful model called Mythos 5 — same underlying architecture, but wrapped in safety guardrails. It topped benchmarks in coding, science, and long-horizon tasks. Developers were excited. Subscribers got it included in their paid plans.

Then, three days later, it was gone.

On June 12, at 5:21 PM Eastern, the US Commerce Department issued an export control directive — basically a national security order — telling Anthropic to cut off access to Fable 5 and Mythos 5 for any foreign national, anywhere in the world, including Anthropic's own non-citizen employees. Since Anthropic had no way to instantly sort US citizens from foreign nationals in its user base, it did the only thing it could: it pulled both models for everyone, worldwide.

Now here's where the story gets messier. There are two very different versions of why this happened.

The government's version, delivered publicly by David Sacks — a senior White House tech adviser — is that the administration warned Anthropic that Fable 5 had been jailbroken (meaning someone found a workaround to bypass its safety filters). The government asked Anthropic to either fix the vulnerability or take the model down. According to Sacks, CEO Dario Amodei refused both options. The ban, in their telling, was a direct consequence of Anthropic's defiance — not government overreach.

Anthropics version is quite different. The company says the directive letter it received included zero specific technical details. Anthropic's own read of the situation is that the jailbreak in question was narrow and minor — the kind of thing that affects other publicly available models too. They complied with the order immediately while objecting to it, calling it disproportionate and lacking transparency. A source close to the company disputes that Anthropic was ever presented with details or refused to fix anything.

So you've got a he-said/she-said between one of the most prominent AI companies in the world and the Trump administration, playing out in real time while millions of users are locked out of a product they paid for.

The geopolitical layer adds another wrinkle. One part of the story involves SK Telecom — South Korea's biggest wireless carrier, a $100 million investor in Anthropic, and a partner in a classified AI project called Project Glasswing. The White House flagged SK Telecom as having suspected ties to China, and asked Anthropic to revoke their access to Mythos first. Anthropic complied. Then Amazon researchers separately found and reported potential vulnerabilities in Fable 5 to the White House. That's what apparently triggered the broader ban.

For users, the practical fallout is immediate: Anthropic opened a refund window for affected subscribers, with a hard deadline of 11:59 PM on June 20, 2026 — which is today. If you upgraded your plan between June 9 and June 14 expecting Fable 5 access, you may be eligible. One catch: if you subscribed through the Apple App Store, Anthropic says it can't process your refund directly — you'll need to go through Apple.

As of now, no restoration date has been confirmed. Anthropic has filed a legal challenge, but that could take months or longer to resolve. President Trump told reporters that negotiations are 'going well,' but no official announcement has been made. In the meantime, the model that was ranked #1 on at least one major coding benchmark remains offline.