OpenAI launches embedded assessments with external labs
According to OpenAI, the company is partnering with Apollo Research, METR Evals, Transluce AI, and Goodfire AI to run embedded assessments that grant deep access to internal models for alignment investigations and safeguard testing.
SourceAnalysis
OpenAI announced on November 16, 2026, that it is working with ApolloResearch, METR_Evals, TransluceAI, and GoodfireAI on embedded assessments that grant outside experts deep access to internal models and systems.
The initiative focuses on investigating alignment and testing safeguards as an extension of OpenAI's ongoing alignment and safety research, according to the company's statement on X, to bring independent scrutiny to risk identification and management in its systems.
OpenAI
@OpenAILeading AI research organization developing transformative technologies like ChatGPT while pursuing beneficial artificial general intelligence.