Cybersecurity
GitHub Pilot Turns AI Security Findings Into 55 Open Source Fixes

Key PointsAI triage and human review cut 9,791 raw findings down to 99 private reports.
View more
- AI triage and human review cut 9,791 raw findings down to 99 private reports.
- Maintainers kept control over which findings to fix and when to ship patches.
- Expanding the pilot depends on expert review capacity that GitHub admits is not yet scalable.
GitHub reported on November 17, 2026, that a security pilot with Microsoft produced 55 fixes across six open source projects as of November 15, 2026. The Codename MDASH trial filtered 9,791 possible issues into 99 private reports, each checked by researchers and accompanied by a proof of concept and suggested fix.
Fourteen of the fixes had already been included in releases. The projects were AutoGPT, Caddy, Express, FastMCP, ImageMagick and Netty. Maintainers decided which findings to fix, how to fix them and when to ship, with reports and patches staying on GitHub.
Human Review Narrowed the Scan Results
The design addresses a maintenance problem created by cheaper AI scanning: duplicates and reports that are not actual vulnerabilities already consume maintainer time, while remediation still requires human judgment. GitHub wants researchers to consult maintainers before pointing another model at their projects and to submit patches alongside any findings.
Microsoft’s Codename MDASH system scanned each project with its MAI Cyber model and suggested fixes. A GitHub Security Lab AI triage agent, built on tailored Copilot skills, removed duplicates, checked validity and focused on high-impact bugs. Seven Microsoft FORGE Lab researchers then manually checked every remaining finding before reports went to maintainers through GitHub private vulnerability reporting.
Maintainers Used the Suggested Patches
The first fixes shipped about a week after the initial scan, according to GitHub. Caddy’s maintainers fixed all 22 reports they received in a week or less. ImageMagick, maintained by two people, also received 22 reports; in one case, its maintainers pushed a fix in 15 minutes after a summary identified the exact lines of code and showed that an apparently familiar issue was different.
At AutoGPT, a report arrived three days after maintainers shipped a broken check and pointed to the right line. They took the fix further than the suggested patch, wrote it within a day and had it in production less than a week later. Across the pilot, maintainers used some suggested fixes as written and adapted others.
A Netty maintainer described the remaining constraint: “Finding vulnerabilities is no longer our bottleneck. Fixing them is the hard part, and that depends on maintainers.”
Expansion Depends on Review Capacity
GitHub described the project sample as small, the conclusions as early lessons and the manual review as not yet scalable; researchers spent about two days per project checking findings by hand. Low-quality reports still reach maintainers every day.
GitHub also identified private-fork CI/CD testing and maintainer access to cyber-capable models as remaining needs. The tests would verify fixes on private forks, and model access would help maintainers verify a proof of concept before and after remediation. Its recommended standard for AI-generated reports is to check each finding against the project’s existing knowledge, threat model and security policy, report privately, include a suggested fix and let maintainers set the pace.
GitHub plans to bring these lessons into private vulnerability reporting and security advisories. It will offer AI security scans with findings checked before private reporting, plus GitHub Copilot Max, to all alumni projects of the GitHub Secure Open Source Fund and extend the support to future Fund sessions. GitHub and Microsoft are also working to expand researcher review while preserving those checks.
For AI-assisted vulnerability disclosure, the pilot’s trade-off is concrete: automation can reduce a flood of raw findings to a manageable private queue, while expert review and maintainer control determine whether those reports lead to patches. Expansion hinges on preserving that standard without returning the triage burden to maintainers.












