GitHub and Microsoft AI pilot helped fix 55 open-source flaws
Original title: GitHub and Microsoft Launch AI Security Initiative to Protect Six Major Open Source Projects
Scans were narrowed to 99 private reports with suggested fixes; 14 fixes had shipped in releases by November 15.
GitHub and Microsoft said on November 17 that an AI-assisted pilot helped maintainers fix 55 vulnerabilities across six Secure Open Source Fund projects. Microsoft’s Codename MDASH, running an earlier MAI Cyber model, proposed patches, and a GitHub Security Lab agent plus seven FORGE Lab researchers verified findings before private reporting. Maintainers kept control of what to fix and when to release. GitHub plans to offer verified scans to other fund alums.
Unconfirmed: GitHub says six consenting projects give limited evidence of how the process scales.
- Projects were AutoGPT, Caddy, Express, FastMCP, ImageMagick and Netty.
- Caddy maintainers fixed all 22 reported issues in a week or less.
- ImageMagick’s two-person team got 22 reports and pushed one fix in 15 minutes.
- Researchers spent about two days per project verifying findings.
- Organizations
- GitHub Security Lab
- AI models
- MAI Cyber
From the original article
2 min read at gbhackers.com
GitHub and Microsoft disclosed on November 17 that an AI-assisted security pilot helped maintainers fix 55 vulnerabilities across six widely used open-source projects. The companies narrowed thousands of potential issues to 99 verified reports, all submitted privately with suggested fixes. As of November 15, 14 fixes had shipped in project releases, GitHub announced.
By Divya · gbhackers.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.