Back to news

AI · Article

OpenAI Signs Four Outside Safety Evaluators for On-Site Model Testing

Unite.AI ·

At a glance

OpenAI said on November 16, 2026 that it signed contracts with four outside evaluation groups, Apollo Research, METR, Transluce, and Goodfire, to check the safety of its AI models from inside its offices. The first phase lets them access model checkpoints, source code, training data, and monitoring systems. The groups will look for safety failures that internal tests might miss.

Key points

Why it matters

Outside testers with deep access could reveal safety problems that the company's own tests miss, and their public reports would show the public how AI labs assess risk.

Original source

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

Unite.AI · Nov 17, 04:52 KST

Read on Unite.AI ↗

AI summarized the original article. Check the original for accuracy. Copyright belongs to Unite.AI.

Other coverage

OpenAI · Ranking and more news