Understand this story
On November 17, 2026, OpenAI announced partnerships with four organizations—Apollo Research, METR, TransluceAI, and GoodfireAI—to conduct embedded evaluations, giving external experts deep access to its internal models and systems to investigate alignment issues and test safety safeguards. According to the division of labor announced by OpenAI, METR will investigate misalignment behavior in OpenAI's largest-scale reinforcement learning training to date, including failures that existing tests and monitoring may miss; Apollo Research will assess whether models secretly act against developers' intentions and whether safeguards can detect this; TransluceAI will review the quality and coverage of post-training alignment evaluations and plans to expand the pilot in 2027; GoodfireAI will station interpretability researchers to analyze model internals. OpenAI also said it has preliminarily agreed to conduct broader monitoring evaluations, with the specific scope to be agreed upon by both parties.
AI-generated from public reports
Latest development · Report
OpenAI Launches First Round of Embedded AssessmentsDevelopments
1 confirmed developments
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 17
Greg Brockman
OpenAI Launches First Round of Embedded Assessments
OpenAI announced a partnership with Apollo Research, METR_Evals, TransluceAI, and GoodfireAI to conduct embedded assessments, giving outside experts deep access to its internal models and systems to investigate alignment issues and test safety safeguards. Greg Brockman reposted the news, stating that the first round of assessments has begun.
OpenAIFirst-party
OpenAI Collaborates with GoodfireAI on Interpretability Research
OpenAI announced that GoodfireAI will station interpretability researchers to use its platform to analyze the internal structure of OpenAI models. The related work focuses on training process monitoring, debugging, and alignment, helping OpenAI researchers discover and improve safety. This is part of the embedded evaluation jointly conducted by OpenAI with ApolloResearch, METR_Evals, TransluceAI, and GoodfireAI.
OpenAIFirst-partySelected
OpenAI Conducts Embedded Alignment Evaluations with Four Institutions Including Apollo and METR
OpenAI is collaborating with Apollo Research, METR, TransluceAI, and GoodfireAI on embedded evaluations, giving external experts deep access to its internal models and systems to investigate alignment issues and test safety safeguards. Among them, METR will investigate misaligned behavior in OpenAI's largest reinforcement learning training to date, including failures that existing tests and monitoring may miss; Apollo Research will assess whether models secretly act against developer intent and whether safeguards can detect it; TransluceAI will review the quality and coverage of post-training alignment evaluations and plans to expand the pilot in 2027; GoodfireAI will embed interpretability researchers to analyze model internals. OpenAI also said it has preliminarily agreed to conduct broader monitoring evaluations, with the specific scope to be agreed upon by both parties.
OpenAIFirst-party
OpenAI conducts embedded evaluations with four institutions including ApolloResearch and METR
OpenAI announced a collaboration with ApolloResearch, METR_Evals, TransluceAI, and GoodfireAI to conduct embedded evaluations, giving external experts deep access to its internal models and systems to investigate alignment issues and test safety safeguards. ApolloResearch will evaluate whether models secretly act against developers' intent and whether safeguards can detect this; METR_Evals will investigate misaligned behavior in its largest reinforcement learning run to date, and has preliminarily agreed to conduct broader monitoring evaluations once scope is agreed upon. TransluceAI will evaluate the quality and coverage of post-training alignment evaluations and plans to expand the pilot in 2027, while GoodfireAI will station interpretability researchers to analyze model internals using its platform, focusing on training run monitoring, debugging, and alignment.