Anthropic is pushing to bring its most capable, and most controversial, AI model to a broader audience. The company’s long-term objective is the safe, scaled deployment of Mythos-class models, a move that would put genuinely unprecedented coding and cybersecurity capabilities into far more hands than currently have access.
Claude Mythos Preview, announced on April 7, 2026, is a model that can autonomously discover and exploit zero-day vulnerabilities in operating systems, web browsers, and open-source projects.
What Mythos has already found
Mythos has identified thousands of zero-day vulnerabilities across critical software systems. Among the highlights: a 27-year-old flaw in OpenBSD and a 17-year-old remote code execution issue in FreeBSD.
Rather than releasing Mythos to the public, Anthropic launched Project Glasswing, a controlled initiative that restricts access to a curated list of partners who use the technology defensively to strengthen cybersecurity infrastructure.
