Day 9: Fable 5 and Mythos 5 stay dark as Commerce export order holds

Anthropic's top models are still offline nine days after a Commerce Department directive forced a worldwide pull.

Nine days after a Commerce Department directive landed at Anthropic’s offices at 5:21pm ET on Friday, June 13, Fable 5 and Mythos 5 remain offline worldwide, along with the most capable backend of Claude Code. The order, addressed by Secretary Howard Lutnick to CEO Dario Amodei according to the Wall Street Journal, barred access by any foreign national “whether inside or outside the United States.” That language swept in Anthropic’s own foreign-national staff, which the company says left no operational choice but to disable the models for every customer. Access to other Anthropic models is unaffected.

The sequencing matters. Fable 5 had shipped earlier that same week as the first model in Anthropic’s new Mythos-class tier, which the company said would “exceed those of any model we’ve ever made generally available.” The export-control letter arrived hours into its commercial life.

The administration hasn’t confirmed its reasoning publicly, per TechCrunch, and the letter itself hasn’t been released. Anthropic says the government provided only verbal evidence of a “potential narrow, non-universal jailbreak,” which the company describes as a technique that “essentially consists of asking the model to read a specific codebase and fix any software flaws.” The Wall Street Journal identified the underlying paper’s authors as Amazon security researchers.

That framing has prompted an unusually direct industry pushback. Katie Moussouris, founder of Luta Security, argues the described behavior “cannot meaningfully be fixed, and any attempt would only weaken the model for defense.” An open letter organized by former Facebook CSO Alex Stamos, signed by roughly 100 cybersecurity professionals from Nvidia, Adobe, Zoom, Google, Anaplan and Sophos, contends the same capabilities “can be replicated” in GPT-5.5, Claude Opus 4.8, Sonnet, and Kimi 2.7.

Anthropic’s own response is sharper than its usual register: applying this standard across the industry, the company warned, “would essentially halt all new model deployments.” Nine days in, that’s the unresolved question sitting under the silence. A frontier release was switched off by a Friday-evening letter whose reasoning hasn’t been shown, and the precedent is now load-bearing.