At 3:28 p.m. Pacific on Friday, November 17, 2023, the board of OpenAI published a 304-word statement on the company's website. The chief executive, Sam Altman, had been removed. The board had concluded he was “not consistently candid in his communications with the board, hindering its ability to exercise its responsibilities.” The board no longer had confidence in his leadership. Greg Brockman, the company's president and a board member, was being removed as chair but would remain at the company. Mira Murati, the chief technology officer, would serve as interim chief executive.
Within hours Brockman had resigned. By Saturday, The Information reported, three senior researchers had also quit. Late on Sunday night, Microsoft chief Satya Nadella announced that Altman and Brockman would lead a new advanced AI research group inside Microsoft. On Monday a letter to the board circulated among the staff, and by the end of the day 738 of OpenAI's roughly 770 employees, more than 95 percent, had signed it. The letter demanded the board's own resignation and Altman's reinstatement, threatening to follow him to Microsoft if neither happened. By Tuesday, November 21, an agreement had been reached. Altman would return as chief executive. The existing board would be substantially reconstituted. Helen Toner and Tasha McCauley would depart. Adam D'Angelo would remain as the only continuing director.
The episode lasted five days. It remains the best test of how the AI industry, when forced to choose between governance and execution capability, actually chooses.
Altman, then 38, had been chief executive since 2019. He co-founded OpenAI in 2015 with Elon Musk and others as a nonprofit. He oversaw the 2019 conversion to a capped-profit subsidiary controlled by the original nonprofit board. Before OpenAI he ran Y Combinator from 2014 to 2019. The board that fired him had six members: Altman himself, Brockman, chief scientist Ilya Sutskever, Quora chief executive D'Angelo, RAND scholar McCauley, and Georgetown Center for Security and Emerging Technology (CSET) researcher Toner. The four-vote majority that removed Altman was Sutskever, D'Angelo, McCauley and Toner.
The board never publicly elaborated on the candor finding. The closest thing to an authorized account came in May 2024, when Toner appeared on the TED AI Show podcast and gave several specific examples. The board had learned of ChatGPT's launch on November 30, 2022, she said, by reading about it on Twitter rather than being briefed in advance. Altman had not disclosed his ownership of the OpenAI Startup Fund, an investment vehicle that bore the company's name and used limited partner capital from Microsoft and others; the fund was structured such that Altman personally controlled it. There was, in Toner's account, a longer pattern of the chief executive providing the board with information that was technically accurate but materially incomplete. “On multiple occasions, he gave us inaccurate information about the small number of formal safety processes that the company did have in place,” she told the podcast.
McCauley gave her account jointly with Toner. In an essay in The Economist on May 26, 2024, the two former directors defended the decision to remove Altman, citing senior leaders who had told the board he fostered a culture of lying. D'Angelo has not spoken publicly about the firing. Sutskever, in a tweet posted Monday, November 20, three days into the crisis, wrote: “I deeply regret my participation in the board's actions. I never intended to harm OpenAI.” He gave a fuller account under oath on October 1, 2025, in a deposition in Elon Musk's lawsuit against Altman and OpenAI, released that November. He described a 52-page memo he had written for the independent directors, which said Altman exhibited “a consistent pattern of lying.”
A review by the law firm WilmerHale, commissioned by the reconstituted board and completed March 8, 2024, was the official institutional answer. The review's published summary concluded that the prior board had acted within its broad discretion to terminate Altman, but that his conduct did not mandate removal. The decision, the summary said, “did not arise out of concerns regarding product safety or security, the pace of development, OpenAI's finances, or its statements to investors, customers, or business partners.” Rather, the firing arose from a breakdown in trust between the prior board and Altman, and the board had acted without a full inquiry or an opportunity for Altman to address its concerns. The reconstituted board reaffirmed Altman's leadership.
Microsoft's role in the resolution is the part of the episode that is most often understated. Nadella's Sunday announcement of an Altman-led research group inside Microsoft was, on its face, a contingency plan. In structural terms it was a credible threat that, absent restoration, the company's largest investor would simply absorb the chief executive, the senior researchers, and through them the practical capacity of the lab. Microsoft had committed roughly $13 billion to OpenAI through preferred-return arrangements inside the capped-profit. The investment was structured around the assumption that the operating company would continue to exist. The Sunday announcement made clear that Microsoft was prepared to redefine what continuing meant.
The staff letter was the second pressure point. Its near-unanimity was its message. The letter did not argue that the board's decision was substantively wrong. It argued that the staff's professional judgment about the company's leadership was at variance with the board's, and that absent reconciliation the staff would leave. The signatures included Sutskever's. The board's own chief scientist, who had voted to remove Altman three days earlier, had by Monday signed a letter demanding his return.
What happened in the weeks and months after restoration is the part of the episode that lasted. The board was reconstituted with directors more conventional in profile, including former Treasury Secretary Larry Summers and former Salesforce co-CEO Bret Taylor, who became chair. The WilmerHale review went forward and reported in March. By the end of 2024 OpenAI was telling employees and the public that its corporate structure “must evolve.” On May 5, 2025, Altman announced the conversion of the operating company to a public benefit corporation, with the original nonprofit retaining control. By October 28, 2025, the recapitalization was complete. The renamed OpenAI Foundation sat above the operating company, now called OpenAI Group PBC. The structure that the November 2023 board had been trying to enforce was gone. The structure that replaced it was closer in form to Anthropic's, which had been a public benefit corporation since founding.
A board that loses its chief executive and accepts him back has either learned something or surrendered. The WilmerHale review's official answer was that the board had acted within its discretion, but that Altman's conduct did not mandate removal. The structural answer is that Microsoft's investment, the staff's loyalty, and the absence of any frontier alternative made restoration the only stable equilibrium available. The board could not credibly run the company without Altman. Microsoft would not let its $13 billion walk into the open market. The staff would not work for whoever the board appointed in his place. There was no second OpenAI to defect to that was not Altman's, because OpenAI's frontier position rested in part on Altman's specific ability to assemble the capital, the talent and the public posture that produced it. The board's authority on paper was real. The board's authority in practice required Altman's voluntary submission to it, which after Friday afternoon was no longer available.
Whether the board was right on the substance of the candor finding remains, nearly three years on, unresolved. Toner and McCauley have defended the decision in public, and Sutskever has described his case against Altman under oath. D'Angelo's continuation on the reconstituted board matters: the only original director who voted to remove Altman and remained afterward has not spoken publicly about the firing since. The WilmerHale review's framing allows both readings to coexist, which is why observers still read the episode according to how they weigh governance against execution.
The institutional lesson of November 2023 is that any governance structure that cannot survive the loss of its chief executive does not actually constrain him, and any chief executive who cannot be replaced is, regardless of formal title, the institution. The 2025 PBC conversion was the formal acknowledgment of what the November 2023 episode had already settled. The five days in November had shown that the board could remove the chief executive on paper but could not keep him out.