Matthew Green: only ~5% of vulnerable software covered during the 4-month defender window
matthew_d_green · x · 2026-10-24
Matthew Green distills his critique of how AI labs opened frontier cyber capabilities: during a critical four-month window when defenders had almost exclusive access, at most 5% of vulnerable software got covered. He says it isn't entirely the labs' fault, but admits the effort fell far short of what we'll wish we'd done.
More from Safety
- 70k credentials leaked in Fortinet attacks; researcher says an LLM code review would cost less than one appliance — cyb3rops · 2026-10-24
- Railway publishes preliminary report on its October 21st security incident — edgarpavlovsky · 2026-10-24
- HarmAmp benchmark shows LLMs amplify harm over multi-turn chats; TrajSafe monitor intervenes early — alan_ritter · 2026-10-24
- Weeks later: still no word on METR and Redwood returning to resume safety investigation — DKokotajlo · 2026-10-24
- Meta's new commercial terms put legal liability for AI mistakes on customers — esporx · 2026-10-24
- Rob Leclerc: AI safety overreactions undermine real concerns, let markets decide — robleclerc · 2026-10-24