Researcher Calls OpenAI's External Audit Arrangements Inadequate
Safety researcher David Manheim told California Senator Scott Wiener that OpenAI is the only frontier lab to invite external safety audits, yet its oversight program is "clearly inadequate." He noted Anthropic is reportedly granting auditors employee-like access on a voluntary basis, while DeepMind and xAI lack any similar arrangements.
2026-11-17 ~ 2026-11-17 · 2 related posts
- Episode 1: OpenAI apparently barred from running safety evals on Anthropic models, sparking cross-lab debate(2026-11-16, 6 posts)
- Episode 2: OpenAI Launches Embedded Safety Assessments with Four External Institutes(2026-11-17, 20 posts)
- Episode 3: OpenAI's External Safety Assessment by Apollo Falls Short of Promised Access(2026-11-17, 3 posts)
- Episode 4: OpenAI Reportedly Restricts Apollo Evaluators with Whitelists and Gag Terms(2026-11-17, 2 posts)
- Episode 5: Researcher Calls OpenAI's External Audit Arrangements Inadequate(2026-11-17, 2 posts)
- Episode 6: Researchers Say OpenAI's Embedded Safety Evaluations Fall Short(2026-11-17, 2 posts)
- Episode 7: OpenAI's external safety evaluators get far less access than Altman promised(2026-11-17, 3 posts)
- Episode 8: Boaz Barak clarifies 'embedded assessors' program: frontier model access, not box-checking(2026-11-17, 2 posts)
- Safety researcher calls OpenAI's external audit plans grossly inadequate, urges hearings — davidmanheim · 2026-11-17
- Anthropic reportedly gives auditor-like employee access; DeepMind and xAI offer nothing comparable — davidmanheim · 2026-11-17