Voices
Anthropic Mythos Breached via Discord Group — Third-Party Contractor Guessed Preview URL
A Discord group gained unauthorized access to Anthropic's Mythos cybersecurity model on its announcement day by leveraging a third-party contractor's access and guessing the model's location based on previously leaked knowledge from AI training startup Mercor. Anthropic says no evidence of impact beyond the vendor environment. Mythos, which finds zero-day vulnerabilities and is restricted to 40 organizations under Project Glasswing, is the most sensitive AI model to be breached to date.
Source
↳ Follow the thread