EventCybersecurityPolicy
Jun 24, 3:29 AM
Anthropic's Mythos model found vulnerabilities in classified US systems
During a testing exercise with US intelligence agencies, Anthropic's Mythos model identified vulnerabilities in classified systems within hours. Sen. Mark Warner disclosed the finding, though it remains unclear if the model exploited the vulnerabilities. The testing was conducted under Anthropic's Project Glasswing initiative.
Jun 24, 3:29 AM