Gary Marcus blasts Anthropic cofounder's AI-safety defense of METR report redactions
GaryMarcus · x · 2026-11-06
Gary Marcus shared and mocked Anthropic cofounder Jack Clark's defense of the heavy redactions in METR's risk report. In the quoted post, the cofounder admits much was redacted — making it harder for others to verify METR's work — but argues that revealing exactly what drives the speedup is commercially sensitive and risks handing others a roadmap to go faster, worsening the very risks METR warns about.
Marcus and the original poster criticize this as rationalizing any decision under the banner of AI safety, sacrificing external verifiability.
Related event: Anthropic Redacted Most Key Evidence from METR's Public Evaluation Report(3 posts)→
More from AGI Musings
- Ethan Mollick: an AI stock downturn won't derail the AI trajectory — emollick · 2026-11-06
- Do NDAs for frontier AI info reduce transparency? Debate over embedded evaluators — EigenGender · 2026-11-06
- Model value as conditional entropy removed: why rising world entropy hurts model companies — xiaosun86 · 2026-11-06
- The most valuable secrets in AI are no-go theorems, not new tricks — fkasummer · 2026-11-06
- GitHub commits hit 275M/week as AI-generated tokens commoditize frontier labs' edge — DimitrisPapail · 2026-11-06
- Argument: AGI's value lies in the long tail, not a few frontier pockets — kevinsurace · 2026-11-06