OpenAI’s former CTO Mira Murati just testified under oath that CEO Sam Altman didn’t tell the truth to her, and that his habits interfered with her ability to do her job. The allegation in question, that Altman lied about safety practices, was already public, having featured heavily in a recent New Yorker feature about Altman. Now it’s court testimony.
For a few weird days in 2023, Murati was an interim CEO of OpenAI after Altman was briefly fired (And then OpenAI very briefly picked another CEO, Emmett Shear). Reports from this chaotic time paint a picture of a working relationship between Altman and Murati that came under strain when Murati lost Altman’s trust. Behind-the-scenes accounts say she sent memos to the company’s board of directors and Altman himself, questioning Altman’s managerial abilities, and that his termination soon followed.
On Wednesday, a video deposition from Murati was shown to the court during the Musk v. Altman court proceedings, in which she testified to the truth of this story. A new model was being prepared for release—which the New Yorker says was GPT-4 Turbo—and in her testimony, as described by the Verge, she said Altman told her OpenAI’s legal department, headed at the time by Jason Kwon, said it wasn’t necessary for the OpenAI safety board to review the model.
The person asking questions in the deposition asked, “As you understand it, was Mr. Altman telling the truth when he made that statement to you?”
To which Murati said, “No.” She later explained in her testimony, according to the Verge, “I confirmed that what Jason was saying and what Sam was saying were not the same thing.” She characterized this as a “misalignment” between Altman and Kwon. Kwon is now OpenAI’s chief strategy officer.
According to Reuters, Murati said in her testimony, “My concern was about Sam saying one thing to one person and completely the opposite to another person.” She also reportedly testified that Altman had been “creating chaos.” She described OpenAI at this time as “at catastrophic risk of falling apart,” and said she was “concerned about the company completely blowing up,” according to Reuters.
According to the New Yorker, Murati’s memos to Altman and the board came soon after this interaction, and “Soon afterward, the board made its decision to fire Altman.” Murati was interim CEO for a handful of days, then Shear stepped in for a few days, and then Altman was reinstated, a move Murati publicly supported.
According to Forbes, Murati testified Wednesday that after his return to OpenAI, Altman continued with behaviors that had worried her, including delays around important decisions, and giving inconsistent messages to different coworkers, which she reportedly said created a “very difficult and chaotic environment.”
About ten months after Sam Altman was reinstated as CEO, Murati left, and a few month later, founded her own AI company.
