Morning Minute: Claude Mythos Breaks Post-Quantum Cryptography

Anthropic reports that its Claude Mythos Preview model identified and demonstrated exploits for vulnerabilities in widely used cryptographic software. Under Project Glasswing, the AI analyzed libraries implementing TLS, AES-GCM, and SSH, potentially enabling certificate forgery or communication decryption. This research illustrates how frontier AI reshapes defensive vulnerability research across browsers, operating systems, and network services. While enhancing defensive capabilities, these autonomous identification and exploitation skills also present governance challenges regarding potential offensive cyber activity.