Report: Anthropic's Restricted Mythos Model Used to Probe US Government Code for Flaws

Anthropic's restricted AI model "Claude Mythos" is reportedly being used by a top-tier US cyber agency to hunt for vulnerabilities in government code. An intelligence agency used Anthropic's Mythos model and found flaws while testing sensitive government systems, an anonymous US government official said, according to Federal News Network. The report says that, through the Anthropic-led public-private initiative "Project Glasswing," vulnerabilities were identified within hours in some cases.