Overview
- The Commerce Department ordered Anthropic on June 12 to block foreign‑person access to its most advanced models, Fable 5 and Mythos 5, citing national‑security concerns over exploit‑finding capabilities.
- Anthropic disabled both models worldwide the same day because it said it could not reliably enforce nationality checks on cloud access without cutting service for all users.
- Government officials and some reports said Mythos breached classified systems during testing, but follow‑up coverage has said that claim referred to a controlled red‑team evaluation and remains contested.
- The Five Eyes cyber and intelligence chiefs issued a joint warning on June 22 that frontier AI will reshape cyber risk within months, urging faster patching, reduced internet exposure, and wider use of defensive AI.
- The dispute raises three immediate tensions: defenders lose powerful tools used to find and fix bugs, companies face regulatory and commercial uncertainty, and allies are pressing for trusted‑partner access or domestic alternatives.