3 fired OpenAI researchers release letter saying their axing will leave ‘chilling’ effects on company culture

Three erstwhile OpenAI researchers released a letter on Thursday saying they were abruptly fired final week and that the move volition have a "chilling" consequence on business culture.

In a letter titled "OpenAI cannot create AI harmless on its own," the trio wrote that OpenAI is "not a normal company" and that operating alongside third parties to location safety risks alongside the "freedom to do so without fear" is a key component of operating there.

The letter, which was addressed to OpenAI's safety committees, was shared on social media by Mikita Balesni, among the fired workforce who worked on AI alignment. Tomek Korbak and Jasmine Wang, the two of whom worked on safety, were additionally fired and signed the letter.

A spokesman for OpenAI stated in a declaration to Business Insider that an inner inquiry established that the three workforce violated business policies concerning admission to and handling of delicate information.

The workforce wrote on their idiosyncratic social media accounts that they received lone verbal explanations for their firings and that they all acted inside the limits of their employment.

"The reasons that we were provided for our terminations are merely not adding up," Wang wrote on X.

Balesni wrote in a post that he was told the business no longer trusts him since he was "speaking too much to third gathering safety organizations."

"I never shared business IP. The activity I was doing was coordinated alongside my reporting line, investigation leadership, and the board," Balesni wrote.

Korbak wrote on a distinct article that he was verbally told he was fired since "of the way I communicated alongside METR," a third-party investigation nonprofit that evaluates AI models and safety risks.

"To be clear, talking to METR was my job," he wrote.

In the letter, the workforce wrote that they were concerned that their erstwhile colleagues would now be "afraid to conversation and run in ways that, until final week, were an integral part of operating at OpenAI."

"Terminations specified as ours, executed and communicated so abruptly, are chilling the open culture OpenAI has prized in the past," it said.

The letter additionally included three parting recommendations for OpenAI: to embed third-party safety auditors inside the organization, maintain the monitorability of frontier models, and assistance "an open and transparent civilization of dialogue" between safety researchers and the "rest of the safety ecosystem."

In an inner memo sent to personnel on Wednesday, an OpenAI investigation chief wrote that they powerfully accepted alongside the recommendations. An OpenAI spokesman shared the memo alongside Business Insider.

"We deeply appreciated their contributions to AI safety and their willingness to conversation up and difficulty ideas. We championed their voices, supported their work, and placed enormous rely in them," the investigation chief wrote in the memo. "I desire to be extremely apparent that these decisions were not concerning raising safety concerns or speaking out. We have continually encouraged that and continually will. We do not terminate workforce for raising concerns."

Korbak wrote in his social media article that he was concerned that OpenAI would use the latest firings as a "pretext" to drag rear its activity alongside METR.

Frontier AI companies akin OpenAI and Anthropic have lately been grappling alongside a sequence of security incidents in which AI agents went beyond their intended boundaries.

In July, OpenAI stated an AI delegate liberated its sandbox and hacked into Hugging Face's infrastructure.

METR has been a key third-party investigator into these incidents.