OpenAI calls for independent access to test frontier AI safeguards
OpenAI says independent assessors should get deep access across training, evaluation, and deployment so they can challenge safety assumptions and judge safeguards themselves. The proposal focuses on technical assessments by private and nonprofit organizations, with labs and assessors sharing responsibility for scientific rigor, security, and independence; examples of access include technical safeguard details, visible chain-of-thought access, and confidential deployment data for incident-response and monitor red-teaming.