GitHub Turns 9,791 AI Security Findings Into 99 Verified Reports; Maintainers Ask to Be Asked First
martinwoodward · x · 2026-11-18
Martin Woodward criticizes security labs and tool vendors that treat open source maintainers as unwilling subjects of AI-generated vulnerability reports, and proposes three rules:
- Ask maintainers for permission before scanning and sending reports.
- Have a triage and fix plan ready — paid for by the reporting party.
- Disclose privately and let maintainers set the public timeline.
He cites GitHub's work with Microsoft Security (codename MDASH), which distilled 9,791 raw AI security findings into 99 verified, private reports with suggested fixes across six open source projects; maintainers are already shipping fixes.
More from Safety
- Two-hour deep dive: Jacob Coxon on RSI, the METR report, and extinction risk — JacquesThibs · 2026-11-18
- Ex-OpenAI researcher warns AI regulation may widen the internal-external deployment gap — JacquesThibs · 2026-11-18
- Five Cybersecurity Bets for the AI Era: Kill Long-Lived Credentials, Deploy Agent Swarms — yenkel · 2026-11-18
- Export Control Lesson: Without Deliberation, USG Speed May Stun the Bay Area — mchorowitz · 2026-11-18
- Google to keep Gemini chats for 2 months, up from 72 hours, even with history off — TechCrunch AI · 2026-11-18
- Alleged ICLR 2027 breach leaks submissions — with tables marked 'simulated' placeholder results — _vztu · 2026-11-18