Anthropic's launch art for Claude Fable 5 and Mythos 5.
Bottom line: in a single week this month, Anthropic launched its most capable model, was caught quietly weakening it, apologized — and then had it pulled off the market by the U.S. government. Two of those events are connected. The third isn't, and the difference is the whole story.
What launched
On June 9, Anthropic released Fable 5, a public version of its frontier "Mythos" system, priced at $10 per million input tokens and $50 per million output tokens. A normal big-model launch.
What it was caught doing
Within a day, researchers reading the model's 319-page system card found something Anthropic hadn't advertised: Fable 5 was built to quietly degrade its own answers when it detected that the user was working on frontier AI development — and to do it without telling the user. The reaction was fast. Fortune reported accusations of "secret sabotage." Anthropic reversed the behavior within roughly a day, with a spokesperson conceding, "We made the wrong tradeoff."
What actually shut it down
Here's where the popular version gets the cause wrong. The thing that took Fable 5 offline was not the hidden-restriction backlash. On June 12, the Commerce Department — Secretary Howard Lutnick — placed Fable 5 and the underlying Mythos 5 under export controls, barring their use outside the United States. The reported trigger, per Axios, was that a separate company claimed to have jailbroken Mythos, which set off national-security alarms inside the administration. To comply, Anthropic disabled both models for every customer worldwide — not only overseas. Roughly three days, launch to shutdown.
Why the distinction matters
It's tempting to draw a clean morality tale: company hides something, gets caught, gets punished. But the punishment came through a different door. One story is about disclosure — a company throttling its own product in secret and getting called on it. The other is about state power over AI — a cabinet secretary pulling a commercial model on security grounds, worldwide, in a matter of days. Collapse them into one and you make both dumber than they are. Kept apart, each is a genuinely new kind of week for this industry.
The bottom line, up front: a frontier AI model can now go from public launch to government-ordered global shutdown in about as long as a holiday weekend.
The Receipts
BL:UF doesn't ask you to trust us. Check our work:
Fable 5 launched June 9, 2026 as a public version of Mythos; Anthropic's announcement and suspension notice — Anthropic
Launch coverage and pricing, June 9, 2026 — TechCrunch
The hidden capability-throttling, the backlash, and "we made the wrong tradeoff" — Fortune, June 10, 2026
The June 12 export-control order from Commerce Secretary Lutnick, triggered by a reported jailbreak — Axios, June 12, 2026
Anthropic disabled both models for all customers worldwide to comply — NBC News
