Industry change
OpenAI starts first phase of embedded outside assessments
- Published
- Source
- OpenAI Blog
Summary
OpenAI announced signed agreements with four independent assessors—Apollo Research, METR, Transluce, and Goodfire—for the first phase of embedded assessments. The assessors get deep access to internal resources such as model checkpoints, training and evaluation data, monitoring systems, and interviews with researchers to independently examine alignment and safeguards. The post also details on-site work, security rules, oversight by the Safety and Security Committee, and dispute-resolution steps, framing the program as part of its commitments under the White House Accord on Super Intelligence.
Why it matters for our work
Evaluation is moving beyond outside testing, with assessors working inside the developer's own systems. How effective it is will only be clear once public reports arrive, but it marks a notable first shift in how the AI industry verifies and oversees frontier development.
Translated from the Korean original. Summaries may be translated and edited. Commentary reflects our perspective; forecasts remain the source’s views.