Anthropic launches Claude Mythos Preview for autonomous hacking

Claude Mythos represents a "watershed moment" for AI-driven security where models can now autonomously find and exploit zero-day vulnerabilities across multiple platforms.

• –Autonomously discovered a 27-year-old bug in OpenBSD and complex exploit chains for web browser renderer escapes.
• –Non-expert engineers were able to use the model to develop sophisticated exploits in under 24 hours.
• –Project Glasswing's restricted release model sets a precedent for how frontier AI with extreme cyber-offensive potential is managed.
• –The model's ability to reverse engineer closed-source binaries effectively ends the "security through obscurity" era for proprietary software.
• –Highlights an emergent reasoning capability where general intelligence gains translate directly into offensive cybersecurity power.