www.securityweek.com 5/25/2026, 11:11:46 AM · external

Anthropic’s AI spots thousands of flaws in open‑source code

Anthropic’s AI spots thousands of flaws in open‑source code
CyberSIXT Evidence Panel Source marked as original reporting
Primary Source anthropic.com

ANTHROPIC'S Claude Mythos model has identified over 23,000 potential vulnerabilities in more than 1,000 open source software (OSS) projects. Out of these, 1,726 vulnerabilities have been confirmed, including over 1,000 rated as 'high' or 'critical'. With ongoing scans, estimates suggest that this number could reach 6,200. Despite a low number of patches released so far, Anthropic expects more to come as vendors respond within the 90-day Coordinated Vulnerability Disclosure policy window.

Major organizations like Mozilla and Palo Alto Networks have reported significant findings attributed to Mythos, while concerns remain over the potential misuse of this AI tool. Anthropic is also working on enhancing safeguards to prevent such misuse, while widening access through Project Glasswing.

View full article

Article by CyberSIXT

Timeline Coverage

Swipe to explore timeline