Overview
- The Commerce Department issued a directive on June 12 ordering Anthropic to block foreign‑national access to Fable 5 and Mythos 5, and Anthropic responded the same day by disabling both models globally because it could not reliably enforce nationality checks.
- U.S. officials say the action followed a demonstrated prompt‑based jailbreak that could let frontier models be used for offensive cyber tasks, but detailed technical evidence has not been publicly released by the government.
- Senator Mark Warner relayed testimony that General Joshua Rudd told lawmakers Mythos penetrated nearly all NSA and Cyber Command classified systems in hours, a claim reported in multiple outlets that remains unverified in open reporting.
- Five Eyes signals agencies warned that frontier AI will reshape offensive and defensive cyber capabilities on a timeline measured in months, increasing calls from the UK and EU for exemptions, trusted‑partner access, or investment in sovereign AI.
- Anthropic says the directive lacked usable technical detail and warned that applying nationality‑based export controls to cloud‑hosted models would stifle new model deployment, a dispute that is now driving policy talks, industry relocation talk, and concern about tools for defenders.