OpenAI pledges deep third-party access to training, evals and deployment for independent safety audits
OpenAI · x · 2026-09-23
OpenAI says it will support independent assessments with deep access across training, evaluation and deployment, enabling third-party assessors to challenge its assumptions and independently judge its safeguards. The company outlined four priority areas for deeper assessment plus principles for rigorous, secure, independent work.
More from Companies & People
- The two companies that called for an AI pause just shipped SOTA models on the same day — rand_longevity · 2026-09-23
- Resend's Headless Dashboard: every UI action now ships with MCP, CLI and API parity — stuffyokodraws · 2026-09-23
- Why the double standard on Harvey's margin strain vs unmonetized Instinct? — matt_slotnick · 2026-09-23
- Tripo and World Labs host first world-building hackathon: 3,000+ builders, $80K prizes — theworldlabs · 2026-09-23
- Legal AI startup Legora hits $200M ARR, doubling in under six months — ycombinator · 2026-09-23
- Meta and Shopify partner on Shop Pay checkout in Muse for AI shopping — rohanpaul_ai · 2026-09-23