GitHub says AI security pilot produced 99 checked reports and 55 fixes

GitHub and Microsoft reported on November 17 that an AI-assisted security pilot produced 99 privately submitted vulnerability reports across six open-source projects, with maintainers fixing 55 by November 15. Fourteen fixes had reached a release, according to GitHub's account.
The pilot used Microsoft's Codename MDASH scanning system and its earlier MAI Cyber model, not the newly announced MAI Cyber 2. It initially flagged 9,791 possible issues. A GitHub Security Lab triage agent filtered findings before seven Microsoft FORGE Lab researchers reviewed the remaining cases manually.
AutoGPT, Caddy, Express, FastMCP, ImageMagick and Netty participated with their maintainers' agreement. Reports included a proof of concept and suggested fix; maintainers retained control over what to change and when.
The review process still required roughly two days per project from the researchers, so GitHub cautions that the workflow does not yet scale easily. It plans to offer checked scans to every Secure Open Source Fund alumni project. The results highlight a practical distinction for AI security tools: producing more alerts is not the same as delivering verified, actionable repairs.
Sources and citations
Revision history
Published .