Anthropic’s Mythos AI broke into almost all NSA classified systems in hours

Senate testimony reportedly claims Anthropic’s Mythos AI breached nearly all NSA and US Cyber Command classified systems within hours, leading the US government to force Anthropic to restrict or shut down its most capable models for non-US users; these allegations remain unverified, lack technical details or indicators, and the article also highlights that Mythos has been used to uncover numerous vulnerabilities and raises geopolitical concerns about access to frontier AI for defensive and offensive cybersecurity purposes.