When OpenAI announced Sam Altman’s return to the helm on Tuesday evening, an impromptu party broke out at the company’s San Francisco offices.
As staff celebrated the reinstatement of the co-founder who was dramatically sacked just four days earlier, investors in the company and other tech leaders rushed to offer their support to Altman, with a flood of heart emojis washing across social media site X.
“Congrats to OpenAI! Great to see the good guys win!” wrote Salesforce founder Marc Benioff. He was one of a number of influential figures in Silicon Valley’s tight-knit tech scene to offer public support, even though he had tried to poach OpenAI’s employees earlier in the crisis, with an offer to match their salaries.
Altman’s return was the culmination of days of feverish efforts, conducted largely remotely, by staff, financial backers and high-profile Silicon Valley figures to persuade a resolute board that they had made a catastrophic error in sacking the 38-year-old.
A mass revolt by OpenAI staff was the primary reason for the board’s about-face, according to multiple people with direct knowledge of the discussions. Many had expected Altman to return to the company on Sunday, encouraged by the appearance of their former boss in the office.
Altman made clear his reinstatement would require an overhaul of the board that had ousted him days earlier for not being “consistently candid”. They included Ilya Sutskever, a co-founder whose work focused on artificial intelligence research; Adam D’Angelo, chief executive of question-and-answer service Quora; technology entrepreneur Tasha McCauley; and Helen Toner from the Center for Security and Emerging Technology at Georgetown University.
The directors initially held firm, appointing Emmett Shear, the co-founder of video-streaming service Twitch, as interim chief executive on Sunday night.
When Sutskever invited employees to an all-hands meeting to greet the new boss on the company’s internal messaging system, he received a stream of middle-finger emojis in response, according to two people with knowledge of the messages.
“The majority of the individuals in the building at that point left,” said one of the people.
Altman swiftly confirmed he and Greg Brockman, another co-founder who left on Friday, would join Microsoft, OpenAI’s key partner. Former colleagues would have an open door to follow and join a new AI unit, said Microsoft chief Satya Nadella.
By Monday morning, the threat of a staff exodus from OpenAI was clear: more than 500 of the company’s 770-strong workforce — including Sutskever — put their names on a letter calling on the board to resign and reinstate Altman. By Tuesday evening, all but 20 people on OpenAI’s payroll had signed.
Staff efforts to bring back Altman were led by a trio of remaining senior executives — Mira Murati, Brad Lightcap and Jason Kwon.
At the same time, OpenAI’s venture capitalist backers had sprung into action, pledging public support to Altman, whatever he did next.
Privately, investors and executives at the company pressed the board for more details on what had prompted Altman’s sacking, beyond its cryptic statement implying he had not always been candid. They were given no further details, according to multiple people with knowledge of those discussions.
According to venture capitalist Vinod Khosla, an early OpenAI backer, Sutskever had raised concerns about the rapid advancement of generative AI’s capabilities.
However, one person with direct knowledge of the negotiations with Altman said the board’s decision was not based on concerns about research progress at the company or any type of malfeasance. Instead, it reflected a loss of trust between the board and Altman due to several instances of obfuscation, they said.
Feeling that the company — and with it, their prospect of billions of dollars in investment returns — was in jeopardy, some investors started exploring legal action, according to two people with knowledge of their thinking. One question was how to sue a board whose duty under OpenAI’s charter was to the safety of humanity at large, not to investors.
Nonetheless, “the threat of the lawsuit was taken seriously”, said one employee at a venture firm invested in OpenAI. “The company didn’t want that. It helped them come to the table and try and find a real solution here. It got serious.”
Among those who rallied to Altman’s side were Airbnb founder Brian Chesky, prominent venture investor Ron Conway and former Salesforce executive Bret Taylor, according to people with knowledge of the situation. Taylor has since been named chair of the new board.
While Sutskever had switched sides, the three remaining board directors — D’Angelo, Toner and McCauley — worked together to negotiate the future of OpenAI as a united team, according to one person close to their thinking.
Before announcing Shear’s appointment on Sunday, board members had explored alternative options for chief executive. One name discussed was Dario Amodei, a former OpenAI employee who left to set up rival company Anthropic, according to a person with knowledge of the discussions.
The prospects of Amodei rejoining OpenAI or a potential merger of the companies were raised, but discussions did not get beyond preliminary stages, said the person. A person with knowledge of the board’s position over the weekend denied that the board had approached Anthropic to discuss any deal.
Anthropic declined to comment.
The board consistently found itself on the losing side of the public argument raging about OpenAI’s future. Altman’s “superpower is getting people onside, shaping narratives, pushing situations into the shape that works for him”, said one person with direct knowledge of the negotiations between Altman and the OpenAI board.
The board “likely underestimated” the strength of Altman’s “PR campaign”, and were probably also “legally constrained by what they could, or were willing, to say”, said another person who has worked with two of the board members. One person with direct knowledge of the board’s position confirmed they were legally constrained during the negotiations.
By Tuesday, with the US Thanksgiving holiday looming, discussions among a small core group — which included Altman, flanked by his longstanding friend Chesky, other OpenAI leadership figures and the board — had entered their fifth day.
D’Angelo, Toner and McCauley negotiated jointly, according to two people with knowledge of the matter, and secured an independent investigation into the events of the past week, as well as assurances that Altman would not rejoin the board of OpenAI.
Shear also played a key role in the talks, one person said.
One sticking point was the new board’s composition. Apart from the remaining directors being keen to ensure that Altman was not on it, the CEO’s backers realised it was important to select people whom the investor community would take seriously but would be regarded as sufficiently independent, said one person with knowledge of the talks.
OpenAI announced on Tuesday that D’Angelo, Taylor of Salesforce and former US Treasury secretary Larry Summers would comprise the new board.
Selecting a third person — in the end Summers — had been particularly difficult, said one person familiar with the talks, and a long list of potential candidates had been floated as options.
The two people who left the board, McCauley and Toner, were both AI safety experts seen as relative outsiders, while remaining director D’Angelo is an established figure in Silicon Valley.
Soon after 10pm on Tuesday night in San Francisco, OpenAI announced a deal had been reached and Altman would be returning. “These last 5 days, I saw people across OpenAI remaining calm and resolute in driving their mission despite all that was happening around them,” Nadella said on X.
Alfred Lin, a partner at venture capital firm Sequoia Capital, which has invested in OpenAI, said he was “inspired” by Altman and Brockman’s “unwavering optimism and commitment to their mission & impact”.
On Wednesday, Sutskever — one of the directors who voted to oust Altman — said nothing could convey “how happy I am” at the outcome of the talks.