Anthropic Project Glasswing Finds 10,000 High and Critical Bugs in 1 Month

Anthropic's Project Glasswing, a collaborative cybersecurity initiative, has identified more than 10,000 high or critical-severity software vulnerabilities in its first month of operation. The project uses the company's advanced Claude Mythos Preview AI model to scan codebases for security flaws across critical infrastructure, cloud platforms, and open-source software.

Approximately 50 partner organizations, including major technology firms and security providers, participated in the effort. Several reported dramatic increases in bug discovery rates. Cloudflare said it found 2,000 bugs, with 400 rated high or critical severity, and noted the model's false-positive rate was better than that of human testers. Mozilla reported finding and fixing 271 vulnerabilities in Firefox 150, a tenfold increase compared to testing with a prior model.

Independent evaluations support the model's capabilities. The UK's AI Security Institute reported that Mythos Preview is the first AI model to fully solve both of its simulated multi-stage cyberattack scenarios. Independent security platform XBOW described it as a significant step up over all prior models, citing unprecedented precision. The model...