Key facts
- Anthropic's AI model, Mythos, identified vulnerabilities in U.S. government computer systems during testing.
- The Mythos model reportedly breached classified systems managed by the NSA and US Cyber Command within hours.
- The U.S. administration restricted access to Anthropic's Fable 5 and Mythos 5 models to U.S. citizens.
- Anthropic subsequently shut down access to these models for all users globally.
- Mythos has also identified thousands of other vulnerabilities, including a long-standing flaw in OpenBSD.
Anthropic's artificial intelligence model, Mythos, has identified vulnerabilities in highly sensitive U.S. government computer systems during a testing exercise conducted in collaboration with U.S. intelligence agencies.
According to a U.S. official, the Mythos model discovered certain vulnerabilities within hours. Senator Mark Warner of Virginia stated that General Joshua Rudd, head of the National Security Agency and U.S. Cyber Command, reported that the AI "broke into almost all of our classified systems, not in weeks but in hours." This testing was part of Anthropic's Project Glasswing initiative, aimed at securing critical software.
Following these findings, the Trump administration directed Anthropic to restrict access to its Fable 5 and Mythos 5 models exclusively to U.S. citizens. Due to the practical difficulties of verifying nationality in real-time, Anthropic chose to shut down access to both models for all users, including international allies and the UK AI Security Institute.
Beyond its success in government systems, Anthropic reports that Mythos Preview has uncovered thousands of vulnerabilities, including a 27-year-old flaw in the OpenBSD operating system. The company has not released Mythos publicly, instead providing access to approximately 200 selected partners.