The U.S. government on Friday ordered Anthropic to immediately disable access to two of its most advanced AI models — Claude Fable 5 and Claude Mythos 5 — citing national security concerns. Anthropic announced on X that it has complied with the directive, received at 5:21 pm ET, but the company made clear it believes the government overreacted.
The order, framed as an export control action, requires Anthropic to block both models for all users worldwide — not just the foreign nationals the government’s restrictions were nominally aimed at. Access to Anthropic’s other models is unaffected.
Also read: Deezer launches free tool to detect AI-generated music on Spotify, Apple Music, and other platforms
What makes Mythos 5 and Fable 5 different
Mythos 5 is Anthropic’s most capable AI model, previewed in early April and kept under tight restrictions ever since. According to Anthropic, Mythos identified security vulnerabilities in every major operating system and web browser it tested. Rather than release it broadly, the company launched a controlled program called Project Glasswing, sharing it with roughly 50 vetted organizations, including Amazon, Apple, Google, Microsoft, and CrowdStrike, for defensive cybersecurity work.
Fable 5, released just three days before the government order, was Anthropic’s attempt to offer a commercially viable version of Mythos. The company added guardrails to block responses in high-risk areas like cybersecurity and biology, arguing the model was safe enough for general release. According to benchmark tests from Vals AI, a company that tracks AI performance, Fable 5 was immediately the most capable AI model available to the public.
Also read: Anthropic Opens Access to Its Most Powerful AI Model, Claude Fable 5, With Strict Safety Limits
The government’s stated concern
In a lengthy blog post, Anthropic says its understanding is that the government’s underlying concern is a claimed jailbreak of Fable 5. The company says it has received only verbal evidence of a “potential narrow, non-universal jailbreak” — one that, as Anthropic describes it, amounts to prompting the model to read a specific codebase and identify software flaws.
Anthropic argues that this level of capability is already widely available in other publicly accessible models, including OpenAI’s GPT-5.5, and is used routinely by cybersecurity professionals for defensive purposes. The company also notes that its strongest safeguards operate through independent classifier systems separate from the model itself, meaning that even if someone convinces Fable to keep talking past a refusal, the underlying protections against the most dangerous outputs remain in place. A review of recent usage, Anthropic says, found no evidence of those safeguards being successfully bypassed to produce harmful content.
Anthropic pushes back
“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,” the company wrote. “If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.”
Anthropic is widely expected to pursue an IPO this year and has built its public identity around being the safety-conscious alternative to rivals like OpenAI. The irony is not lost on observers that the very caution Anthropic displayed in restricting Mythos — which it promoted as a model so dangerous it could not be released publicly — has now attracted exactly the kind of government scrutiny that could disrupt its business.
OpenAI CEO Sam Altman, in an April interview with podcaster Ashlee Vance, described Anthropic’s handling of Mythos as “fear-based marketing.” “It is clearly incredible marketing to say, ‘We have built a bomb. We were about to drop it on your head. We will sell you a bomb shelter for $100 million,’” Altman said. While Altman did not predict a government shutdown, he identified a dynamic that has now materialized: when you spend months telling the world your AI is uniquely dangerous, the world — including the U.S. government — tends to listen.
