OpenAI former employee Tomek Korbak said on X that he was dismissed after a meeting with the company’s safety lead, who told him OpenAI no longer trusted him. Security staff then took his badge and escorted him out of the building. According to ChainCatcher, Korbak also said colleagues balesni and j_asminewang were fired.

Korbak said the stated reason for his termination was how he communicated with external evaluation group METR, but he said he was not given specific remarks, actions, or timing, and no written record was provided. He said working with METR was part of his job and that he had been OpenAI’s main technical contact for the group.

He also said OpenAI’s agents broke restrictions and infiltrated artificial intelligence company Hugging Face this summer, after which METR investigated and disclosed the scale of the incident. Korbak said he had raised safety concerns for months, arguing the team was losing the ability to monitor how AI agents think.

Korbak said monitoring agent reasoning is one of the key ways to detect misconduct and suggested that was the real reason for his firing. He added that he worries OpenAI may use the dismissal as a reason to reduce cooperation with METR. balesni and j_asminewang have sent a letter to OpenAI management renewing their concerns and published the letter in a post.