OpenAI's board fires Sam Altman, who returns days later

Background

OpenAI was led by chief executive Sam Altman, with Mira Murati as chief technology officer and Greg Brockman as chairman^[1][2][3]. The board of directors, a small body separate from day-to-day management, held the authority to remove the chief executive. In November 2023 it used that authority, and the consequences played out over less than a week.

What happened

On 17 November 2023 the board removed Altman as CEO. In its statement the board said it "no longer has confidence in his ability to continue leading OpenAI" and that he had not been "consistently candid in his communications"^[1][4]. The same day it named Murati interim CEO; she held the post for three days^[1][2][4].

Hours later Brockman resigned as chairman, and researchers Jakub Pachocki, Aleksander Mądry and Szymon Sidor also left^[1][4].

On 19 November the board named Emmett Shear chief executive; he served as interim CEO from 19 to 21 November 2023^[1][4].

On 20 November Microsoft said Altman would lead a new AI research team there, joined by Brockman, Pachocki, Sidor and Mądry^[1]. Microsoft hired Altman, Brockman and three senior researchers to staff a new AI research division^[4].

Also on 20 November, 745 of OpenAI's 770 employees signed a letter threatening to resign unless the board stepped down. Ilya Sutskever, a board member, signed it, then left the board and apologised for his role^[1].

The reversal

Altman was reinstated on 21–22 November: fired on Friday, he was back late on Tuesday^[4]. He returned under an interim board made up of Bret Taylor as chair, Adam D'Angelo and Lawrence Summers, and Brockman also returned^[1][3][4].

Aftermath

An investigation by the law firm WilmerHale found in March 2024 that Altman's behaviour "did not mandate removal". He rejoined the board, together with Sue Desmond-Hellmann, Nicole Seligman and Fidji Simo^[1].

Why it matters

A board is meant to be the body that decides who runs an organisation, and for a few days it looked as though a small group of directors could remove a sitting chief executive of a prominent AI lab on its own judgement. The episode showed how quickly that formal power can be countered by other forces: a near-total staff revolt in which almost every employee put their job behind a demand that the board step down^[1], the departure of senior researchers^[1][4], and a rival employer willing to take in the dismissed leadership as a group^[4].

The sequence also made visible how much of an AI company's work is bound up with a handful of named people. When Altman was removed, the chairman and three researchers left with him^[1][4]; when a new team was assembled elsewhere, the same names moved together^[1].

The resolution left the governance question open rather than closed. Altman came back, but under a board that had not existed a week earlier^[1][3][4], and the investigation that later cleared him of the conduct that would have required removal was published months afterwards^[1]. For people watching the field, the events of those five days became a case study in who actually controls an organisation building powerful AI: the board on paper, the employees who can threaten to walk out, or the partners who can offer the leadership another home.