Project Glasswing: When AI Becomes the Ultimate Hacker—and Defender

This podcast/blog post covers Anthropic's Project Glasswing and its unreleased AI model Claude Mythos, which is reported to be capable of finding zero-day vulnerabilities, generating exploits, and assisting with fixes. The piece discusses the implications for cybersecurity, red teaming, and AI-driven threats and notes that the model is considered too powerful for public release.