She sat under oath and, for the first time in public record, turned a newsroom anecdote into a legal admission. I watched the video deposition and felt the room shift—what had been gossip now had the weight of testimony. You can sense how a private memo becomes a public turning point.
I’ve followed OpenAI’s internal dramas for years; you’ve probably read the New Yorker piece, the Reuters summaries, and The Verge’s blow-by-blow. What changed on Wednesday was simple and sharp: Mira Murati, OpenAI’s former CTO and brief interim CEO, said on camera that Sam Altman wasn’t telling her the truth about how the company handled model safety reviews.
In a conference room where a new model was being prepped, someone said the safety board didn’t need to sign off
That line—about a model reportedly GPT-4 Turbo—was the spark. Murati testified that Altman told her OpenAI’s legal team, led then by Jason Kwon, had said the safety board’s review wasn’t necessary. When asked under oath if Altman was telling the truth, she answered No.
I want you to feel the friction: two senior leaders giving different directives, and the person in charge of technical safety left holding the ambiguity. Murati told the court she confirmed Kwon’s version did not match Altman’s, and she called the mismatch a “misalignment.”
Did Mira Murati testify that Sam Altman lied?
Yes. In a deposition shown during the Musk v. Altman trial, Murati said Altman told her one thing and said something else to others. She said she confirmed that Jason Kwon’s statements and Altman’s statements were not the same, and she answered directly—under oath—that Altman had not been truthful in that instance.
At a whiteboard, someone scribbled memos and then handed them to the board
Murati says she wrote memos to the board and to Altman, questioning his management as chaos spread. Reports from late 2023 describe a rapid, almost theatrical sequence: Altman was fired, Murati briefly became interim CEO, Emmett Shear took over for a few days, and then Altman was rehired. I’ve read those memos; they pushed a private worry into a formal record.
She told the court she feared the company was “at catastrophic risk of falling apart” and worried it might “completely blow up.” Those are not casual phrases; they framed the board’s later decision, according to the New Yorker’s reporting.
What did Murati say in her deposition about Altman’s behavior?
She described inconsistent messaging, stalled decisions, and behavior she said created a “very difficult and chaotic environment.” Reuters reports she accused Altman of “creating chaos” and said he would tell different people different things—an allegation that, once sworn into testimony, becomes part of an official record rather than newsroom color.
In one hallway conversation she feared trust had been lost
The relationship frayed. Behind closed doors, Murati says she lost trust in Altman’s leadership. That loss is what the New Yorker and Bloomberg traced to the memos and to the board’s decision to fire him, even if his reinstatement followed within days. I’ve seen executives recover trust; I’ve also seen leadership collapse when contradictory signals spread. At that moment, OpenAI felt like a house of cards.
After Altman returned, Murati told the court his pattern of mixed messages and delayed decisions continued. Forbes covered her testimony that these behaviors persisted and worsened internal morale. About ten months after Altman’s reinstatement, Murati left OpenAI and later founded her own AI company.
At a deposition table, testimony converts rumor into record
This is the legal calculus: what used to be a magazine anecdote now sits under oath and in court transcripts. The Verge, Reuters, Bloomberg, and others have taken that media narrative and watched it harden into evidence. For anyone watching OpenAI—investors, engineers, regulators—the shift matters. It’s no longer chatter; it’s sworn testimony.
You should also note the players: Jason Kwon, who was head of legal during the episode and is now OpenAI’s chief strategy officer; Emmett Shear and Sam Altman; the New Yorker, The Verge, Reuters, Forbes, and Bloomberg as the public record. These names and outlets are the scaffolding that turns internal conflict into a story with consequences.
Why did Murati send memos to the board?
She told investigators she believed Altman’s conflicting messages and decision delays put the company at risk. The memos, according to the New Yorker, were her attempt to put a technical leader’s safety concerns onto the board’s agenda. When leaders disagree in private, a memo can be the only reliable timestamp.
What happens next is the question that will keep the tech world scanning headlines and court filings. I’ve tracked crises that began with emails and ended with leadership exits; I’ve seen reputations rebuilt when transparency follows. Right now, the record is public, the witnesses are named, and the allegation—Altman telling different stories to different people—has been sworn to. It’s like a conductor dropping his baton: the music doesn’t stop immediately, but everyone notices the change in tempo.
Do you think a single deposition can change how regulators, investors, and engineers view OpenAI going forward?