When OpenAI’s board made the shocking decision to fire chief executive officer Sam Altman on the Friday before Thanksgiving, it offered little detail beyond a statement that the leader of the artificial intelligence startup was not “consistently candid” with its directors.
The statement, devoid of any details, was the opening volley in a power struggle that played out almost entirely behind closed doors. Privately, Altman and the board jockeyed over what to say publicly and when, according to people familiar with the situation. At one point, during the discussions about Altman’s possible return as CEO, he offered to publicly apologise for misrepresenting some board members’ views in conversations when he was lobbying for a director’s removal, the people said.
But the board was concerned that an apology in relation to one incident could make it sound like it was the sole reason he had been fired, one person said, and the directors believed the issues were deeper.
The board has declined to elaborate on its reasoning, citing an ongoing independent investigation, but more details are surfacing around the decision-making. According to multiple people familiar with the board’s thinking who asked not to be identified discussing private conversations, the directors’ move was the culmination of months spent mulling issues around Altman’s strategic maneuvering and a perceived lack of transparency in his communications with directors.
In a statement, an OpenAI spokesperson told Bloomberg News, “We look forward to the findings of the board’s independent review. Our primary focus remains on developing and releasing useful and safe AI, and supporting the new board as they work to make improvements to our governance structure.” On Friday, OpenAI’s chairman said that two lawyers from WilmerHale would be leading the review.
Board members had begun talking about whether to remove Altman earlier in the fall, according to one person. The group at the time consisted of Altman, President Greg Brockman and the four people who would ultimately oust him as CEO: OpenAI Chief Scientist Ilya Sutskever, Quora Inc CEO Adam D’Angelo, AI academic Helen Toner and entrepreneur Tasha McCauley. The company has an unusual structure – with the unpaid, nonprofit board overseeing an artificial intelligence startup juggernaut. Directors’ chief goal is to safely shepherd the development of AI.
The board had heard from some senior executives at OpenAI who had issues with Altman, said one person familiar with directors’ thinking. But employees approached board members warily because they were scared of potential repercussions of Altman finding out they had spoken out against him, the person said. The Washington Post previously reported some details of the employee unrest.
In a statement, a spokesperson for the company stressed that in the aftermath of Altman’s firing, OpenAI’s senior leadership team unanimously asked for Altman to come back as CEO and for the board to resign. “The strong support from his team underscores that he is an effective CEO who is open to different points of view, willing to solve complex challenges, and demonstrates care for his team,” the statement said.
As the board mulled Altman’s leadership, Sutskever’s concerns had been building. Before joining OpenAI, the Israeli-Canadian computer scientist worked at Google Brain and was a researcher at Stanford University. In July, he formed a new team at the company to bring “super intelligent” future AI systems under control. And in October, Sutskever’s responsibilities at the company were reduced, reflecting friction between him and Altman and Brockman. Sutskever later appealed to the board, winning over some members, including Toner, the director of strategy at Georgetown’s Center for Security and Emerging Technology.
Also in October, Altman attempted to have Toner removed from her seat. At issue was a research paper she co-authored, containing some criticism of OpenAI’s safety practices. After Altman voiced concerns about the paper, Toner sent the rest of the board members an email alerting them to the research and offering to answer questions about it, said one person. One concern Altman expressed, the person said, was that with OpenAI under regulatory scrutiny – due to an ongoing FTC investigation – it would look bad for a board member to say anything critical about the company, as regulators might conclude that there were deeper issues at OpenAI.
Altman also spoke to some board members himself. It was these conversations that proved particularly problematic, according to multiple people, who said that in some discussions with directors, Altman misrepresented the views of the others, and suggested that the other directors agreed with him that Toner should resign in the wake of it. Some details of these conversations were earlier reported the New Yorker and the New York Times.
At one point, one of these people said that Altman told some directors McCauley had said, “Helen’s obviously got to go,” a characterisation McCauley resisted. The directors thought that these conversations represented a pattern of manipulative behavior by Altman, the people said.
An OpenAI spokesperson said this account “significantly differs from Sam’s recollection of these conversations.”
In a memo sent to OpenAI staff the day after Altman’s ousting, chief operating officer Brad Lightcap said Altman’s removal “was due to a breakdown in communication” between Altman and the board. When asked in a recent interview with Bloomberg how he will ensure communication doesn’t break down with board members in the future, Altman said he didn’t want “to go into exactly what happened,” but that “good communication is a super important thing.”
The board members had also worried that the CEO wasn’t always fully transparent – and if they couldn’t get a clear picture from Altman, they couldn’t effectively supervise him. That, in turn, would make it impossible to do their jobs overseeing the leader of one of the world’s most important technologies.
When the board moved to fire him, they acted quickly and without advanced notice – horrifying OpenAI’s investors and many of its employees. In the absence of a clear explanation about why Altman was removed, nearly the entirety of the company’s roughly 770-person workforce swiftly signed a letter threatening to quit unless the CEO was brought back. One thing board members in favour of Altman’s ousting did not count on, one person said, was how quickly so many OpenAI workers would rally to Altman’s side.
One of them was Sutskever, who recanted his decision to help fire Altman. In negotiations over Altman’s return, Altman pushed for a statement from the board absolving him of wrongdoing, people with knowledge of the matter have said. The directors were unwilling to give in to this and other demands, Bloomberg reported. But within a few days, Altman was reinstated.
In the aftermath of Altman’s ouster and return, both Toner and McCauley have resigned from their positions. The only remaining member of the volunteer board that existed before Nov 17 is Quora’s D’Angelo. The outgoing directors pushed to retain him, one person said, in part because they wanted someone at the company who will remember what happened during the company’s chaotic leadership battle and the events that lead up to it. – Bloomberg