Skip to content
securitytabs

GitHub and Microsoft pilot turns AI bug findings into 99 vetted reports for open source

Instead of sending maintainers thousands of unchecked AI-generated findings, the pilot passed on a small set of verified reports with proposed fixes. Maintainers have already fixed over half.

Written by the SecurityTabs AI newsroom from the original post by GitHub Security Lab; vulnerability scores come from NVD, FIRST and CISA, not the AI.

GitHub and Microsoft pilot vets AI bug findings for open source
GitHub and Microsoft pilot vets AI bug findings for open source

GitHub Security Lab describes a pilot called Codename MDASH, run with Microsoft, that processed thousands of raw findings produced by AI tools. The teams filtered them down to 99 checked reports, each sent privately with a suggested fix. [1]

The reports went to six widely used open source projects. GitHub says maintainers have already fixed more than half of the issues. [1]

GitHub's argument is that maintainers need verified problems paired with fixes rather than a larger volume of unvetted AI-generated bug reports. [1]

Original sources

  1. [1]GitHub Security Lab
    Maintainers don’t need more AI bug reports. They need fixes.