OpenAI is finalizing contracts with outside safety assessors and will announce details in the coming weeks, reaffirming its commitment to embedding independent evaluators. It also agreed with the researchers that preserving the ability to monitor advanced models requires an industrywide commitment. Their letter urged AI companies to halt work that would reduce humans' ability to monitor AI reasoning and to cooperate with outside auditors.
Jasmine Wang, Tomek Korbak and Mikita Balesni were fired the previous week and dispute OpenAI's account. Korbak said he was dismissed over communications with METR, although serving as its technical contact during an investigation of the Hugging Face incident was his job. Wang said she accidentally opened an executive's sensitive email through access granted for recruiting that IT had failed to remove, and reported it within minutes. Balesni said he believed they were fired for “prioritizing safety over the near-term interests of OpenAI as a corporation.”