Anthropic has unveiled OSS Scanner, a new service that provides free, AI-powered security scans to open-source projects. The company says participating projects will receive thorough, periodic vulnerability assessments powered by its strongest models, including the recently released Claude Mythos.
The initiative addresses a persistent gap in open-source security. Many critical libraries and frameworks that underpin modern software rely on volunteer maintainers who lack the resources for dedicated security audits. By offering automated scanning at no cost, Anthropic hopes to surface vulnerabilities earlier in the development cycle, before they can be exploited at scale.
OSS Scanner works on an opt-in basis. Project maintainers who enroll grant Anthropic's models access to their codebase for analysis. The resulting reports flag potential security issues, ranging from common coding mistakes to more subtle architectural weaknesses. Anthropic says the reports are designed to be actionable, with specific remediation guidance rather than generic warnings.
The move positions Anthropic alongside other AI labs that have invested in open-source security. However, the company acknowledges a trade-off: the reports generated by OSS Scanner will not be made public, meaning the broader community cannot benefit from the findings unless project maintainers choose to disclose them.
Industry observers note that the service also serves as a practical demonstration of Claude Mythos's capabilities in code analysis and security reasoning, areas where AI models have shown rapidly improving performance over the past year.