Anthropic vs. US government: the timeline
Three escalations in five months
US government forces Anthropic to pull Fable 5 and Mythos 5 for all users worldwide
The Commerce Department issued an export control directive on June 12 ordering Anthropic to block all foreign nationals from Fable 5 and Mythos 5. With no way to verify user nationality at scale, Anthropic shut off both models for every customer globally — the first time a deployed frontier AI model has been taken offline by US government order. Anthropic complied but publicly disputed the legal standard, calling the cited jailbreak narrow and comparable to vulnerabilities in other deployed models.
The Commerce Department ordered Anthropic to cut off all foreign nationals from Fable 5 and Mythos 5 on Friday evening, and Anthropic had no practical way to comply without taking both models down for everyone. By late Friday, Claude's landing page read: "Fable 5 is temporarily unavailable." 1
This is the first time the US government has forced a leading AI company to pull a commercially deployed model offline.
What the order says
The directive, signed by Commerce Secretary Howard Lutnick and drafted with help from the Bureau of Industry and Security (BIS), cited national security authorities to suspend access to Fable 5 and Mythos 5 by "any foreign national, whether inside or outside the United States, including foreign national Anthropic employees." 3
Anthropic received the letter at 5:21 PM ET on Friday. No specific details of the national security concern were included.
The government's stated concern: a narrow method to bypass Fable 5's cybersecurity safeguards — a "jailbreak." Anthropic reviewed the demonstration and concluded it could find only "a small number of previously known, minor vulnerabilities," all of which are accessible using other publicly available models, including OpenAI's GPT-5.5. 1
Why Anthropic shut off all users, not just foreign nationals
Anthropic's infrastructure has no reliable way to verify the nationality of every active user. The only compliant path was to disable Fable 5 and Mythos 5 for all customers. AWS confirmed late Friday that Anthropic asked it to "revoke access to the models for all users in all regions." 4
All other Claude models remain available. The suspension affects only the two Mythos-class models launched on June 9.
Anthropic's public disagreement
Anthropic complied, but its statement reads as a direct challenge to the legal standard the government applied.
"We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people. If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers." 1
The company called the directive a "misunderstanding" and said it is working to restore access. It also noted that the government provided only "verbal evidence" of the potential jailbreak, and that no testers have found a universal jailbreak — one capable of broadly bypassing Fable 5's safeguards across a wide range of requests.
The policy contradiction this exposes
Anthropic has argued publicly that governments should have the power to block dangerous AI deployments — Dario Amodei made exactly that case at a Bloomberg conference on Wednesday, and the company published a detailed policy framework the same week. 4
Friday's order is the first time that stated position has been tested against an actual government action — and Anthropic's response draws a sharp distinction between the oversight it endorses (transparent, fair, fact-grounded) and the one it just received (no written specifics, verbal-only evidence, broader operational impact than the stated risk).
Dean Ball, a former White House AI adviser who contributed to the administration's 2025 AI Action Plan, wrote on X that the directive effectively means all "non-Americans" are restricted, and that users should expect to prove citizenship to use Anthropic models going forward. 4
Several prominent Anthropic leaders were born outside the United States. An Anthropic spokesperson declined to comment on how the directive applies to them.
統計カードを読み込んでいます…
Where things stand with the government
This is the third major collision between Anthropic and the federal government in five months. In February, the Trump administration put Anthropic on a Pentagon supply-chain blacklist after the company pushed back on letting the military use Claude for domestic surveillance and autonomous weapons. Anthropic sued; a California federal judge ruled in its favor, but the case remains active in a Washington, D.C. court.
Last week, the Financial Times reported that Anthropic engineers were embedded at the NSA helping to deploy Mythos for offensive cyber operations — even as the Pentagon litigation continued. Then, on June 2, Trump signed an executive order on AI that included a voluntary mechanism for the government to gain early access to frontier models. The order seemed to signal an easing of tensions. 3
Friday's export control directive reverses that reading. The Commerce Department, not the Defense Department, is now the source of the action — and the legal instrument is export control, a much broader authority than the supply-chain designation that started the dispute.
For a company weeks away from an S-1 going live, losing access to its two highest-capability models — even temporarily — is a material IPO risk that will require disclosure.
参考ソース
- 1Statement on the US government directive to suspend access to Fable 5 and Mythos 5
- 2Anthropic Walks Back Policy That Could Have 'Sabotaged' AI Researchers Using Claude
- 3Anthropic suspends new AI models after government directive
- 4Anthropic disables most advanced AI models after US order limiting foreign access
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。