According to Jin10, OpenAI announced partnerships with four independent organizations—Apollo Research, Goodfire, METR and Transluce—to launch the first phase of embedded safety assessments. The partnerships will allow external experts to access internal models, source code, training data and monitoring systems to review model alignment and safety measures. Apollo Research will assess whether models covertly act against developers’ intentions; Goodfire will study model mechanisms and training monitoring; METR will examine misaligned behavior during large-scale reinforcement learning; and Transluce will evaluate the quality of safety testing in the post-training phase. Assessors may enter OpenAI offices when necessary to access sensitive materials using dedicated equipment. An independent oversight committee under the OpenAI Foundation’s board will supervise the process and receive full reports. The organizations may publish their findings independently, while OpenAI may request the removal of confidential information or delay disclosure of safety risks. OpenAI also plans to expand the assessments in 2027.