The Trump administration has ordered Anthropic to take Claude Fable 5 and Claude Mythos 5 offline for every user worldwide. The directive, received Friday at 5:21 PM ET, appears to be the first time the US government has forced an AI company to pull a live commercial model over national security concerns.
READ ALSO: Anthropic Raises $65B, Overtakes OpenAI at $965B Valuation
Anthropic complied. In a detailed public blog post, it made clear it believes the government got this wrong, and the consequences for its upcoming IPO could be severe.
What the Government Ordered and Why Anthropic Disputes It
The directive came framed as an export control action, restricting foreign national access to both models. Anthropic had no technical way to restrict access to foreign nationals without cutting off everyone. The shutdown went global for all customers, including Anthropic's own staff.
According to Axios, Commerce Secretary Howard Lutnick wrote to Dario Amodei on Friday, placing both models under export controls covering everywhere outside the US and all foreign nationals inside it. The government did not cite a specific threat in the order itself.
Anthropic says its understanding is that the underlying concern is a claimed jailbreak of Fable 5 specifically, feeding the model a codebase and asking it to surface security flaws. The company calls it a "narrow, non-universal jailbreak" and notes that OpenAI's GPT-5.5 and other publicly available models can already do the same thing. "If this standard was applied across the industry," Anthropic wrote, "it would essentially halt all new model deployments for all frontier model providers."
Why Mythos Made Anthropic a Target
Mythos 5 is Anthropic's most capable model, and the company itself spent months telling the world exactly that. Internal testing showed Mythos could identify critical flaws in every major operating system and browser it analysed. Anthropic chose not to release it publicly. Instead, it launched Project Glasswing, a controlled programme sharing Mythos with roughly 50 vetted organisations among them Amazon, Apple, Google, Microsoft, and CrowdStrike, strictly for defensive cybersecurity work.
READ ALSO: How Difficult Is It to Learn Coding? A Stage-by-Stage Breakdown
Fable 5 had been live for three days when the order arrived. Anthropic built it as a guardrailed version of Mythos; cybersecurity and biology outputs were blocked, and the ceiling was lowered enough for general release. Vals AI, which benchmarks models across providers, placed Fable 5 at the top of its public rankings the moment it launched.
Months of public warnings that Mythos was too dangerous, followed immediately by a guardrailed public launch, that combination drew exactly the scrutiny Anthropic was trying to avoid. In an April interview with podcaster Ashlee Vance, OpenAI CEO Sam Altman said, “It is clearly incredible marketing to say, 'We have built a bomb. We were about to drop it on your head.’ "He did not predict a government order. But he saw the trap, and Anthropic walked into it.
An IPO Filing, a Shutdown, and a Question About What Comes Next: The timing is the story. Less than two weeks before this order, Anthropic filed to go public after closing a $65 billion financing round that valued the company at close to $1 trillion.
READ ALSO: AI Is Creating More Cybersecurity Jobs Than It Kills
Regulatory shutdowns mid-IPO introduce material risk that must be disclosed to investors. They also raise a harder question: can a safety-first company stay at the frontier if the government reads its safety disclosures as evidence of danger?
For developers and businesses worldwide that integrated Fable 5 in the three days it was live, the path forward is unclear. An official told Axios the suspension could last weeks, pending review by the US national security apparatus.
Anthropic wrote in its blog post that governments should be able to block unsafe AI deployments but only through a process that is "transparent, fair, clear, and grounded in technical facts." By its own standard, this action failed every test.
