After months locked behind Project Glasswing, Anthropic announced Claude Mythos will be released to all customers. 93.9% on SWE-bench Verified, 72.4% exploit generation rate --- this is a different league of AI.
Anthropic's most advanced cybersecurity model can autonomously identify and exploit zero-day vulnerabilities. Available only to internet-critical companies and open-source maintainers.
Anthropic's Claude Mythos Preview is a specialized AI designed to discover zero-day vulnerabilities before malicious actors can exploit them. Part of Project Glasswing, it unites over 40 major tech organizations in a proactive security initiative.