So8res: glasswing missing a bug for 6+ months indicts defense-beats-offense theory
harris_edouard · x · 2026-10-22
Replying to Edouard Harris, So8res argues that glasswing failing to find a bug for 6+ months is an indictment of the theory that defense can outpace offense on relevant timescales, regardless of differences between the underlying models. Harris counters that their views on what the theory predicted in April may differ, given the independently trained models' coverage. A substantive AI safety debate on offense-defense asymmetry.
More from Safety
- OpenAI redacted 23 of 61 pages in METR's embedded misalignment-monitor report, researchers say — peterwildeford · 2026-10-23
- After eMudhra fiasco, AI agents set to audit certificates from 23 Indian CAs — kingslyj · 2026-10-23
- Security legend Halvar Flake: Rejecting AI-assisted bug reports over their origin is out of place — mboehme_ · 2026-10-23
- 8B safety classifier trained on 1,275 prompts beats GPT-5.4 across 5 benchmarks — HaggaiMaron · 2026-10-23
- DeepMind's Geoffrey Irving: alignment may need cryptography-grade rigor as AI finds 12 OpenSSL zero-days — geoffreyirving · 2026-10-23
- System safety arguments aren't proofs: lessons from Uber ATG's self-driving case — geoffreyirving · 2026-10-23