US orders Anthropic to disable Claude Fable 5 and Mythos 5 — flagship models pulled
TL;DR
Commerce Secretary Howard Lutnick orders Anthropic to block all foreign users from Fable 5 and Mythos 5 — including its own non-US employees. Anthropic pulls both flagship models globally.
Tonight (June 12) at 5:21pm ET, Commerce Secretary Howard Lutnick sent Dario Amodei an export control order: ban any foreign national from accessing Fable 5 and Mythos 5, in or outside the US — including Anthropic's own non-US employees. Anthropic can't selectively comply, so it pulled both models from all users. Other models, including Opus 4.8, unaffected.
Official reason vs reality: the government claims a company demonstrated a jailbreak technique bypassing Fable 5's guardrails to unlock Mythos's underlying cyber capabilities, constituting a national security risk. Anthropic's reply: «We've reviewed the demonstration the government cited. The capabilities shown also exist in publicly deployed models like OpenAI GPT-5.5, and are routine tools for cybersecurity defenders. If this standard applied to the whole industry, it would essentially halt new deployment of any frontier model.»
This isn't the first hit. July 2025: Anthropic and the Pentagon signed an agreement making Claude the first frontier model approved for classified networks. February 2026: the agreement collapsed when the Pentagon demanded Claude support «all lawful purposes» including autonomous weapons and mass surveillance; Anthropic refused. March 9: Trump administration designated Anthropic a «supply chain risk», banning military and defense contractors from using its models — Anthropic sued in California federal court and on appeal; a judge temporarily blocked the blacklist. June 9: Fable 5 and Mythos 5 released. The administration had tried to block the launch; failing that, it switched to export controls. Three days later, here we are.
Next: Anthropic says more details within 24 hours and is actively pursuing restoration. Commerce indicates the controls «may be lifted in weeks» with separate licensing. Anthropic's IPO is filed confidentially at $965B — the timing of this order is uncomfortable.
via Anthropic
Official reason vs reality: the government claims a company demonstrated a jailbreak technique bypassing Fable 5's guardrails to unlock Mythos's underlying cyber capabilities, constituting a national security risk. Anthropic's reply: «We've reviewed the demonstration the government cited. The capabilities shown also exist in publicly deployed models like OpenAI GPT-5.5, and are routine tools for cybersecurity defenders. If this standard applied to the whole industry, it would essentially halt new deployment of any frontier model.»
This isn't the first hit. July 2025: Anthropic and the Pentagon signed an agreement making Claude the first frontier model approved for classified networks. February 2026: the agreement collapsed when the Pentagon demanded Claude support «all lawful purposes» including autonomous weapons and mass surveillance; Anthropic refused. March 9: Trump administration designated Anthropic a «supply chain risk», banning military and defense contractors from using its models — Anthropic sued in California federal court and on appeal; a judge temporarily blocked the blacklist. June 9: Fable 5 and Mythos 5 released. The administration had tried to block the launch; failing that, it switched to export controls. Three days later, here we are.
Next: Anthropic says more details within 24 hours and is actively pursuing restoration. Commerce indicates the controls «may be lifted in weeks» with separate licensing. Anthropic's IPO is filed confidentially at $965B — the timing of this order is uncomfortable.
via Anthropic
