Skip to content
Unite.AI·Indexed by this site · Originally published· Governance →· Its place in the scan record →SelectedAttention59

OpenAI signs contracts with four independent evaluation organizations to conduct on-site safety evaluations

OpenAI, 상주형 안전성 평가를 맡을 독립 평가기관 4곳과 계약 체결

AI summary

On November 16, 2026, OpenAI announced that it had signed first-phase on-site evaluation contracts with Apollo Research, METR, Transluce, and Goodfire; the four organizations will have access to model checkpoints, source code, training data, and monitoring systems to evaluate alignment and safety guardrails during training and internal use. Apollo Research evaluates whether models covertly act against developer intent and whether guardrails can detect such behavior; Goodfire embeds interpretability researchers to monitor training runs; METR focuses on what OpenAI says is its largest reinforcement learning training run to date, with its initial scope limited to that run; Transluce starts from the quality and breadth of post-training alignment evaluations.

Why it was selected

The original source details each of the four evaluation organizations' evaluation tasks, access scope, and compensation arrangements, which can be used to understand how independent safety evaluations at frontier labs are implemented in practice.

Read originalunite.ai