Anthropic’s Mythos AI model reportedly found vulnerabilities in sensitive U.S. government computer systems during a testing exercise conducted with U.S. intelligence agencies. The results have intensified debate over AI security controls and the Trump administration’s restrictions on Anthropic’s Fable 5 and Mythos 5 models. #Anthropic #Mythos #Fable5 #Mythos5 #ProjectGlasswing
Keypoints
- Anthropic’s Mythos model found vulnerabilities in U.S. government systems during testing.
- The tests were conducted with U.S. intelligence agencies under Project Glasswing.
- Sen. Mark Warner said the tool broke into nearly all classified systems within hours.
- The Trump administration restricted foreign access to Anthropic’s Fable 5 and Mythos 5 models.
- Cybersecurity leaders urged the administration to lift the directive, citing the models’ defensive value.