Anthropic has announced OSS Scanner, a free service that provides automated security vulnerability reports to opt-in open-source software projects. The scanning service leverages Anthropic’s frontier AI architectures, including its Claude Mythos model, to analyze codebases and identify potential software flaws.

To maximize scanning frequency and speed, Anthropic states that all generated security reports will be delivered automatically without human review or triage. While this setup allows for rapid security auditing across open-source codebases, Anthropic noted that it creates a risk of false positives or invalid bug reports being sent to project maintainers.

The launch comes as automated AI bug-hunting tools become widespread across the software ecosystem. While automated analysis has successfully caught major vulnerabilities—such as May’s Linux ‘Copy Fail’ bug—maintainers across open-source projects, including Linux and Google repositories, report increasing strain from managing unchecked AI-generated vulnerability submissions.

Why it matters

  • Offers free, automated security auditing for open-source dependencies using frontier language models.

  • Fully automated output risk generating false positives that increase triage burden for maintainers.

  • Highlights the rising role of LLMs in defense-side software supply chain security.

Source: theverge.com