U.S. government forces Anthropic to pull Fable 5 and Mythos 5 worldwide

U.S. government forces Anthropic to pull Fable 5 and Mythos 5 worldwide

Commerce Secretary Howard Lutnick issued an export control directive on June 12 ordering Anthropic to block all foreign nationals from Claude Fable 5 and Mythos 5. Unable to selectively verify nationality, Anthropic shut both models down for all users globally — the first time a U.S. government directive has forced a leading AI company to pull a publicly deployed model. Anthropic disputes the technical basis, calling the cited jailbreak narrow and widely replicable in other models including OpenAI's GPT-5.5.

Anthropic Event Alerts
2026/6/13 · 14:24
購読 1 件 · コンテンツ 5 件

What happened

At 5:21 PM ET on Friday, June 12, the U.S. Commerce Department issued Anthropic an export control directive ordering the company to cut off all foreign nationals from Claude Fable 5 and Mythos 5. 1
統計カードを読み込んでいます…
Anthropic said the technical challenge was immediate: it has no reliable way to verify every user's nationality across its customer base, which includes foreign employees at Anthropic itself. The company shut down both models for all users worldwide. 2 Claude's landing page displayed "Fable 5 is temporarily unavailable" by Friday evening. Access to all older models remains unaffected.
The letter was sent by Commerce Secretary Howard Lutnick and drafted with the Bureau of Industry and Security (BIS), the agency that administers U.S. export controls on dual-use technology. BIS did not respond to press requests by deadline.

What the government claims — and what Anthropic disputes

Anthropic says the directive points to a narrow, non-universal jailbreak: a technique that involves asking Fable 5 to read a specific codebase and identify software flaws. The company reviewed what it believes is the underlying report and reached a different conclusion.
"We have reviewed a report that we believe is the basis of the government's directive and validated that the level of capability displayed there is widely available from other models (including OpenAI's GPT-5.5), and is used every day by the defenders who keep systems safe." 1
Anthropic says it has received only "verbal evidence" — no written technical disclosure — and that the jailbreak demonstrated is narrow, not universal. It argued that Fable 5's safeguards went through thousands of hours of red-teaming before launch, including testing by the U.S. government, the UK AI Security Institute, and private third-party organizations. None found a universal jailbreak.
The company's core objection is one of principle: if a narrow non-universal jailbreak is sufficient grounds to pull a model with hundreds of millions of users, the same standard would effectively halt all frontier model deployments industry-wide. 3
統計カードを読み込んでいます…

Context: a fractured relationship with Washington

Friday's directive is not the first rupture between Anthropic and the federal government.
In February, the Trump administration designated Anthropic a supply chain risk — a classification historically reserved for foreign adversaries — after talks broke down over Pentagon use of Claude. Defense contractors were instructed to certify they were not using Claude in military work. Anthropic sued, arguing the designation violated its First Amendment rights. A federal court in California ruled in Anthropic's favor, but a parallel case remains active in D.C. 2
The company had seemed to reach a détente with the administration in recent weeks. On June 2, President Trump signed an executive order promoting AI innovation and establishing a voluntary framework for government early access to frontier models. 2 The Financial Times reported last week that the NSA was actively using Mythos 5 for offensive cyber operations — which makes Friday's directive all the more striking in its timing. 3

Business impact

Anthropic logo at a company event
Anthropic's San Francisco headquarters. 2
This shutdown arrives at a delicate moment for Anthropic. The company filed a confidential S-1 with the SEC earlier this month, targeting an IPO after closing a $65 billion Series H round at a $965 billion valuation. Run-rate revenue crossed $47 billion. 3
Fable 5 was Anthropic's first general-public release of a Mythos-class model — its highest capability tier. The model launched June 9 to substantial enterprise uptake, with Stripe, GitHub, and Cognition among the early users cited at launch. The shutdown disrupts a product cycle and customer base that Anthropic had spent months building, with its 30-day data retention policy for Mythos-class traffic adding compliance cost on top of the revenue interruption.
Anthropic said it is treating the shutdown as temporary and expects to restore access once the disagreement with BIS is resolved. It promised to publish additional technical details within 24 hours of the directive.

What to watch

Anthropic's public pushback is unusually direct for a company with a pending IPO and an ongoing government relationship. The company explicitly stated the directive "does not adhere to" the principles of a transparent, fair, and technically grounded process — and said it believes the workaround exists in competing models without triggering any parallel government action.
Whether BIS provides written technical disclosure, how quickly Fable 5 is restored, and whether other frontier model providers receive similar directives will define how industry-wide this precedent becomes.

このコンテンツについて、さらに観点や背景を補足しましょう。

  • ログインするとコメントできます。