METR reveals OpenAI could redact eval reports; researcher calls the terms 'pretty bad'
DKokotajlo · x · 2026-10-23
METR disclosed that in its evaluations of OpenAI, the company could make redactions to reports and METR had no right to unilaterally describe what was redacted. METR says future exercises will provide more robust public transparency via its recent proposal. Anthropic researcher DKokotajlo quoted the post calling the terms "pretty bad" and urging better arrangements in the future.
More from Companies & People
- Nectar, AI-native social OS, raises $40M+ from Menlo, GV and True Ventures — testingcatalog · 2026-10-23
- eMudhra exits SSL business amid botched CA key generation and looming Chrome distrust — saibharadwaj · 2026-10-23
- Every app company claims to be 'a database for agents' — until you look at database pricing — matt_slotnick · 2026-10-23
- A Real Product Strategy Helps Your Team Say No, Says Tim Herbig — aakashgupta · 2026-10-23
- IBM CEO's quantified AI stance vs HPE CEO's 'buzzword soup' — analyst picks a side — DavidLinthicum · 2026-10-23
- CMU professor wins NSF-STRIDE AI Efficiency Challenge grant, hiring PhD students and postdocs — sh_reya · 2026-10-23