Harris vs So8res: can AI-driven cyber defense ever fully match offense?
harris_edouard · x · 2026-10-22
Edouard Harris and So8res debate whether AI defense can outpace offense on relevant timescales. Harris argues the overlap in bug coverage between systems like glasswing and daybreak would only become perfect in a hypothetical endstate where all bugs are found and fixed — which he doesn't believe is achievable in a human economy outside lab conditions — and notes that in April, theory would have predicted non-overlapping coverage for independently trained models.
More from Safety
- 'Ten years hiding in one of the most read codebases': open source's many eyes theory under fire — aronchick · 2026-10-22
- METR red team: two outsiders found ways OpenAI's AI agents could slip past their monitoring system — ross2000 · 2026-10-22
- Composio launches Agent Firewall, an SLM-powered runtime boundary for agent actions — arankomatsuzaki · 2026-10-22
- METR and Redwood embedded at OpenAI for three weeks to assess internal agent misalignment monitoring — CFGeek · 2026-10-22
- Daniel Rock: HFT's market lessons — verification, circuit breakers and an 'immune system' for the internet — danielrock · 2026-10-22
- Security researchers: nation-states already deploying abliterated open-weight models for offense — joshua_saxe · 2026-10-22