Anthropic launches OSS Scanner to give open source projects free AI security scans

Anthropic has launched OSS Scanner, an optional vulnerability detection service for open source software. Core maintainers of eligible projects can apply by sending a pull request to its GitHub repo. The review rules are similar to Google's OSS-Fuzz, focusing on key projects tied to critical infrastructure. Accepted projects get free regular security scans from Anthropic's strongest AI models, including Claude Mythos. Over the past six months, Anthropic used its latest models to scan many core software projects worldwide and found more than 29,000 possible vulnerabilities. Due to limited staff, only about 6,000 got human review. The scan results are fully AI generated, with no human checking. To test an early version, a senior penetration tester reviewed 97 critical and high-risk bugs found in 48 projects. 85 of them (88%) met the disclosure standard, and only one was a false alarm. Anthropic says it will keep improving the system with maintainer feedback and newer models.