At 5:21 p.m. ET on Friday, Anthropic PBC received an emergency Commerce Department directive ordering it to cut foreign nationals off from Fable 5 and Mythos 5, the two AI models it had positioned as the centerpiece of its public-listing pitch. Rather than partition access, the company disabled both worldwide. Fable 5 had launched three days earlier. Vals AI had just rated it the most capable publicly available model in benchmark tests.

The directive, issued under national-security export-control authorities and described by Bloomberg as a Trump administration order, didn’t specify the concern in writing. Officials verbally pointed to a disclosed technique for bypassing safeguards on Fable 5, which Anthropic characterized as a “narrow, non-universal” jailbreak that prompts the model to read a specific codebase and identify software flaws. A demonstration, the company said, surfaced “a small number of previously known, minor vulnerabilities.”

“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,” Anthropic said, arguing the standard, applied industry-wide, “would essentially halt all new model deployments.” OpenAI’s GPT-5.5, comparably deployed, faces no such restriction. Claude Opus 4.8 remains available.

The subtext is harder to miss. In March the Pentagon labeled Anthropic a supply-chain risk, a designation historically reserved for foreign adversaries, requiring contractors to certify they won’t use Claude on military work. Anthropic is challenging that in federal court. David Sacks, the former AI and crypto czar, and Emil Michael, the Pentagon undersecretary for research and engineering, have publicly criticized the company and its executives. Earlier this month it filed confidentially for an IPO at a $965 billion valuation.

Mythos 5, distributed through Project Glasswing to roughly 50 vetted organizations including Amazon, Apple, Google, Microsoft and CrowdStrike, identified flaws in every major operating system and browser it was tested against. Regulators, Anthropic countered, “should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts,” adding that “this action does not adhere to those principles.” Further guidance is expected within 24 hours; Fable’s 30-day data retention runs in the meantime.

Sources