Nathan Lambert: banning open-weight models over cyber risk would be lose-lose
Interconnects (Nathan Lambert) · rss · 2026-10-06
Nathan Lambert argues the open-weights cyber risk discourse is broken: documented attacks come mostly from closed-model APIs, Anthropic's GLM-5.3 report dodges key questions, and banning open weights without banning closed APIs would widen the offense-defense gap. He also explains China's endogenous, registration-based AI safety posture and why Chinese labs release open weights.
More from AGI Musings
- OpenAI reportedly spent tens of millions in compute to crack Navier-Stokes in days — haider1 · 2026-10-06
- Nearly half of DOL-recognized jobs have zero agentic AI tool coverage, Cohere finds — Cohere_Labs · 2026-10-06
- Goertzel's 2010 classic: 'The Singularity Institute's Scary Idea, and why I don't buy it' — burny_tech · 2026-10-06
- Ben Goertzel rebuts Yudkowsky: why the 'Everyone Dies' AGI doom argument is wrong — burny_tech · 2026-10-06
- Will Rinehart walks through AI extinction and catastrophe probabilities: less bleak than you think — WillRinehart · 2026-10-06
- Ben Goertzel's d-calculus: the math of goal preservation under AI self-improvement — burny_tech · 2026-10-06