TREE NEWS update: OpenAI has signed its first embedded evaluation agreements with external expert organizations, including Apollo Research, giving them deep access to its internal models. The experts will investigate model alignment and test OpenAI’s safety guardrails. The company announced the deals on its own channels, framing the move as opening internal model access to outside specialists.
OpenAI Signs First Embedded Evaluation Deals With External Safety Groups Including Apollo Research
The significance here is structural rather than technical: OpenAI is letting outside specialists probe its internal models and guardrails, which shifts safety scrutiny from self-assessment toward third-party evaluation. That matters most for regulators and enterprise buyers who have had to rely on vendor claims, and for other labs facing pressure to match this level of access. Whether embedded evaluation produces findings that are published or acted on, and whether peer labs follow with comparable arrangements, is the open question.
Generated by AI for reference only.
Share on WeChat
Open WeChat → Scan → then tap "…" to send to a chat or Moments.
Tap "…" in the top-right corner to send to a chat or share to Moments.