GitHub and Microsoft announce results of MDASH open source security pilot
Understand this story
GitHub and Microsoft used Microsoft's multi-model agent scanning tool Codename MDASH to conduct a vulnerability scanning pilot on six open source projects including AutoGPT, Caddy, and Express, submitting a total of 99 private vulnerability reports with remediation suggestions. As of November 15, maintainers had fixed 55 of them, and 14 fixes had entered official release versions. During the pilot, seven researchers manually reviewed findings for each project for about two days, and reported them privately only after confirmation. GitHub said it will provide AI security scanning to all projects participating in the pilot, with discovered issues reported privately only after review, and will gift GitHub Copilot Max.
AI overview based on source reports · Updated 2 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Hacker News AIMaintainers don't need more AI bug reports. They need fixes
In a pilot called Codename MDASH, GitHub and Microsoft used Microsoft's multi-model agentic scanning harness to scan six widely used open source projects, then filtered and hand-verified the results, delivering 99 private reports with suggested fixes; maintainers had fixed 55 as of November 15, 14 of them already in a release. The process combined AI scanning, a GitHub Security Lab triage agent, and manual review by seven Microsoft FORGE Lab researchers, with reports sent through GitHub private vulnerability reporting. The post argues that AI-generated findings should be checked, reported privately, and accompanied by fixes, and notes that researcher review does not yet scale.
Activity for this story
Not enough continuous observations to show a trend yet.