During the trial of Elon Musk’s lawsuit against Sam Altman and OpenAI, former OpenAI CTO Mira Murati said in a videotaped testimony that she could not trust then-CEO Sam Altman’s oral statements. She testified under oath that Altman made false statements to her about the safety review process for a new model.
Mulati said in her testimony that Altman told her that OpenAI’s legal department had determined that a new AI model did not need to be submitted to the company’s internal deployment safety board for review. When asked, "To your knowledge, is Mr. Altman telling the truth when he says this?" she answered unequivocally: "No."
Muradi further said that Altman’s behavior during her tenure at OpenAI made her job more difficult, stressing that her criticism was “entirely about management.” She said she was responsible for an extremely difficult job in a highly complex organization. "I was asking Sam to lead, and to lead in a clear way, without undermining my ability to do my job."
The security process surrounding a certain GPT model is a specific case where she and Altman disagreed. After hearing what Altman had to say, Mulati checked specifically with Jason Kwon, who joined OpenAI in 2021 as general counsel at the time and is now chief strategy officer. She said there was a "mismatch" between Kwon's and Altman's statements, and she confirmed that "what Jason said and what Sam said were not the same thing." Out of an abundance of caution, she ultimately still ensured that the model was submitted to a safety committee for review.
This is not the first time Altman has been accused of lying. OpenAI co-founder Ilya Sutskever previously wrote in a 52-page memo to the board that Altman "displays a persistent pattern of lying that undermines his own executives and pits executives against each other." This was read out in another deposition.
Mulati’s testimony appeared in a live update of the “Musk v. Altman” case, providing a rare window into the disagreements within OpenAI’s top brass over security governance and management styles. As more relevant testimonies become public, this lawsuit surrounding the future direction of OpenAI and the behavior of its leadership is further drawing attention from the industry and regulatory agencies to the internal decision-making and safety processes of large AI companies.