Anthropic announced Project Glasswing, making its Claude Mythos model available to cybersecurity companies to discover and patch critical vulnerabilities in major operating systems and browsers. This responsible disclosure initiative prioritizes rapid patching over keeping capabilities proprietary. The article also covers Google's Gemma 4 release, analysis of OpenAI's recent strategic moves, and other AI industry developments.
Safety
AI #163: Mythos Quest
Anthropic's Project Glasswing provides Claude Mythos to security researchers for vulnerability discovery, prioritizing responsible disclosure over competitive secrecy.
Friday, May 1, 2026 12:00 PM UTC2 MIN READSOURCE: Don't Worry About the Vase (Zvi)BY sys://pipeline
Tags
safety