METR: AI labs redacted key claims before approving frontier risk report
kevinnbass · x · 2026-10-17
kevinnbass highlights how frontier labs shape safety evaluations. METR says every company in its May 2026 Frontier Risk Report "exercised its rights to redact or anonymize" key claims before publication. By contrast, SecureBio held veto/redaction rights over its review of Anthropic's risk report but used neither, and Apollo Research states developers cannot redact overall conclusions. A follow-up finds 18 of 43 evaluator organizations have no published conflict-of-interest or publication-control policy, including Faculty, Gray Swan AI, ARC, Irregular and Haize Labs.
More from Safety
- Matthew Green highlights unexplained skipped curves in NIST P-256/P-224 seed generation — matthew_d_green · 2026-10-17
- AI tools reportedly helped one attacker hit 9 South Korean banks — Comfortable_Gene5180 · 2026-10-17
- Meta terms reportedly grant itself rights to scrape the open web for AI training — VeryWellVersed · 2026-10-17
- Jensen Huang on Lex Fridman: build the flywheel, don't just block chip exports — rohanpaul_ai · 2026-10-17
- Macroscopic Ventures: Anthropic backer with stakes in Apollo Research, Gray Swan and Redwood — kevinnbass · 2026-10-17
- Evaluator payment policies compared: METR takes no lab money, Apollo charges market rates — kevinnbass · 2026-10-17