Apollo Embeds Six Evaluators Inside OpenAI to Probe Model Oversight
Apollo Research announced a first-of-its-kind embedded evaluation pilot, stationing six safety evaluators inside OpenAI for four months to investigate whether models attempt to subvert oversight and control.
2026-11-17 ~ 2026-11-17 · 2 related posts
- Apollo Research embeds six evaluators inside OpenAI to probe model scheming — MariusHobbhahn · 2026-11-17
- OpenAI embeds six Apollo Research safety evaluators to investigate AI subversion — RebeccaBellan · 2026-11-17