Anthropic's New Claude Mythos Model Discovers Thousands of Zero-Day Vulnerabilities, Rivaling Top Human Hackers

Anthropic has unveiled a preview of Claude Mythos, a new general-purpose language model with cybersecurity capabilities that match elite human hackers. The model has autonomously discovered thousands of zero-day vulnerabilities across operating systems, browsers, and open-source software. Simultaneously, Anthropic launched Project Glasswing, a global cybersecurity defense initiative bringing together over a dozen tech and financial companies and more than 40 critical infrastructure organizations to accelerate vulnerability patching through early model access.