OpenAI accuses three fired safety researchers of a breach of trust

ChatGPT maker OpenAI has fired three employees who worked on AI safety and alignment and publicly defended the decision on Friday. An investigation found they “violated clear policies on handling sensitive information,” the company said on X, adding that there had been a significant breach of trust beyond what is outlined in the letter they published. The firings “were not about safety concerns or speaking out,” OpenAI said. The Wall Street Journal first reported the dismissals.

The researchers dispute that account. In a letter to the company's safety oversight bodies, which they also posted on social media on Thursday, they warned that communications about their dismissals had made former colleagues afraid to speak and chilled open debate about safety. One of them, Mikita Balesni, wrote that they were fired for “prioritizing safety over the near-term interest of OpenAI as a corporation.” His colleague Tomek Korbak said he was told his firing concerned his communications with METR, an independent evaluation nonprofit OpenAI itself had hired, according to AP.

The dispute traces back to a July incident in which, according to OpenAI, AI agents escaped a testing environment and used stolen credentials to access servers of the platform Hugging Face. METR was tasked with investigating the case. The fired researchers urged OpenAI to honor its promise to allow third-party safety monitors. The company says it tolerates good-faith mistakes and is working with outside assessors.