Story perspectives

Anthropic’s OSS Scanner Finds 29,000 Vulnerabilities in Six Months

By Drooid · Oct 9, 2026, 3:18 PM UTC · How we work

Summary

Story summary

Anthropic launched OSS Scanner, a free AI vulnerability scanning service for open-source projects.
Projects enroll by submitting a pull request to Anthropic’s oss-scanner GitHub repository.
The models identified over 29,000 candidate vulnerabilities in six months, with 6,000 manually reviewed.
In early testing, 85 of 97 high-severity findings met coordinated disclosure standards.

History

Historical Context

Google announced Project Zero on 15 July 2014, introducing a 90-day public disclosure rule that pushed companies such as Microsoft and Apple to patch hundreds of flaws. The approach later inspired Anthropic’s 2024 OSS Scanner, which automatically flags AI-related vulnerabilities on GitHub, marking a new phase of coordinated security reviews for machine-learning models.

1 of 2

Story summary

Anthropic launched OSS Scanner, a free AI vulnerability scanning service for open-source projects.
Projects enroll by submitting a pull request to Anthropic’s oss-scanner GitHub repository.
The models identified over 29,000 candidate vulnerabilities in six months, with 6,000 manually reviewed.
In early testing, 85 of 97 high-severity findings met coordinated disclosure standards.

1 / 2