OpenAI launches embedded assessments with METR and Apollo, but critics say access falls short of promise
BlancheMinerva · x · 2026-11-17
OpenAI announced "embedded assessments" with Apollo Research, METR, Transluce, and Goodfire, giving outside experts deep access to internal models to investigate alignment and test safeguards.
Critics note the access is far narrower than Sam Altman's "employee-like access" pledge two months ago: no Slack, no broad context (OpenAI itself says assessors need freedom to follow evidence—in the HF investigation, assessors were missing 70% of relevant transcripts until they proved incompleteness and got the rest in the final two days), and no confidential info like model weights or architecture, which matters for judging chain-of-thought trustworthiness and understanding how AI is used to build more powerful AI.
Related event: OpenAI Launches Embedded Assessments with Four External Safety Labs(19 posts)→
More from Companies & People
- Google Labs Acquires Howie, an AI Calendar Agent for Small Businesses — jacalulu · 2026-11-17
- Platform Engineering Mentions Surge ~150% as AI Coding Tools Force Governance — craigsdennis · 2026-11-17
- Stanford publishes free 6-lecture CS329Z video series on building AI agents — stanfordnlp · 2026-11-17
- OpenAI says ChatGPT now handles ~20% as many searches as Google — PeterDiamandis · 2026-11-17
- Warp CEO: nearly every team is building an internal 'software factory' for coding agents — thursdai_pod · 2026-11-17
- Profound CTO on why the startup took over every ad in Union Square station — soleio · 2026-11-17