Mar 31, 2025 21:00:00

What happened behind the scenes in the OpenAI CEO Sam Altman's dismissal scandal that shocked the world?

TechCrunch

In November 2023, Sam Altman, co-founder and CEO of OpenAI, the developer of ChatGPT, was dismissed by the board of directors. Altman later returned to his position as CEO , but the internal conflict at the world's leading AI company attracted the attention of many people. Keech Hagee, a reporter for the Wall Street Journal, will publish a book summarizing the behind-the-scenes story of Sam Altman's dismissal , 'The Optimist: Sam Altman, Openai, and the Race to Invent the Future,' on May 20, 2025. An excerpt from the book has been published on the Wall Street Journal's web version.

Exclusive | The Secrets and Misdirection Behind Sam Altman's Firing From OpenAI - WSJ
https://www.wsj.com/tech/ai/the-real-story-behind-sam-altman-firing-from-openai-efd51a5d

OpenAI is a special company consisting of the non-profit corporation 'OpenAI, Inc.' and its subsidiary for-profit corporation 'OpenAI Global, LLC.' The non-profit corporation's mission is to 'build a safe, general-purpose artificial intelligence that benefits all of humanity,' and the board of directors is obligated to the non-profit corporation's mission, not the interests of shareholders.

About a year before Altman's dismissal, OpenAI's board had been stuck in a quandary over who to hire for AI safety. The board had been interviewing Ajeya Kotla, an AI safety expert at the charity Open Philanthropy , but the process had stalled due to opposition from Greg Brockman, who is also a co-founder and board member of Altman.

One of the candidates Altman recommended to the board to replace Kotler was Airbnb CEO Brian Chesky. Chesky told Hagee, 'There was a bit of a power struggle. Basically, whoever Sam named must be loyal to Sam. That's why the other board members said no.'

TechCrunch

At the time of the dismissal scandal, OpenAI's board of directors consisted of six people: Altman, his ally Brockman, co-founder Ilya Sutskever, Quora CEO Adam D'Angelo, entrepreneur and philanthropist Tasha McCauley, and Helen Toner, director of strategy at Georgetown University's Center for Security and Emerging Technology. The latter three have no financial ties to OpenAI.

As OpenAI's AI becomes more and more powerful, the importance of corporate governance and the ability to oversee Altman has become more apparent to some board members. 'Things like ChatGPT and GPT-4 were a meaningful shift for the board to recognize that the stakes are rising here,' Toner said in an interview. Meanwhile, Toner and McCauley were already beginning to lose confidence in Altman.

OpenAI has worked with Microsoft to set up a Joint Safety Committee to assess the risks of new products before they are released. At a Winter 2022 meeting discussing the release of three controversial features of GPT-4, Altman claimed that 'all three features were approved by the Joint Safety Committee,' but when Toner asked for evidence, it turned out that only one had been approved.

Around the same time, Microsoft conducted pre-release testing of GPT-4 in India before it received Joint Safety Committee approval. Altman and Brockman likely knew about this fact, but did not mention it at the board meeting after the release, and the outside directors only found out about it after the six-hour meeting when another OpenAI employee approached them.

The final blow was the revelation at a dinner party held in the summer of 2023 that the OpenAI Startup Fund, an investment fund thought to be run by OpenAI, was actually owned by Altman personally. OpenAI executives explained to the board that Altman established the fund for tax reasons, but this strengthened suspicions about Altman among outside board members.

Then, in late September, Satskever told Tonner to speak with Mira Murati, who was then OpenAI's chief technology officer, and that Altman's 'toxic management style' had been causing problems for years and that Altman's relationship with Brockman was interfering with Murati's work.

Speaking with Satkiver again, Tonner learned that Satkiver had lost confidence in Altman for a variety of reasons. For example, Satkiver had founded a team to pursue the direction of OpenAI's research in 2021, but a few months later, OpenAI researcher Jacob Pachocki founded a similar team and Satkiver's team was absorbed. Satkiver was waiting for the moment when the board would be able to fire Altman, Hagee wrote.

Since then, Satkiver and Murathi have been speaking with board members individually and keeping in daily contact with them. When Toner published

a paper in October that criticized OpenAI's safety, Altman told Satkiver that McCauley had made it clear that Toner should leave the board. However, Satkiver, who had been in contact with McCauley, realized that this was a lie.

Satskever and Murati had been gathering evidence to fire Altman for some time, and at this stage Satskever used Gmail's auto-delete feature to send a PDF document of evidence to Toner, McCauley, and D'Angelo. The document contained dozens of lies and harmful behaviors about Altman, mostly consisting of screenshots of Slack channels taken by Murati. The document also focused on allegations of bullying against Brockman.

On November 16, the board members, excluding Altman and Brockman, met via video call and voted to dismiss Altman. They also decided to remove Brockman from the board, as it would be unlikely that Murati would be able to assume the role of interim CEO if Brockman remained on the board.

On the night of the day Altman's dismissal was decided, Murati, who was attending a meeting, was informed by the board of directors that Altman had been dismissed and that he had been asked to become interim CEO. After Murati was appointed interim CEO, he asked the board of directors why Altman was being fired, but the board did not explain the reason.

On November 17, when the news of Altman's dismissal spread around the world, a meeting was held with OpenAI's board of directors and management. In the lead-up to the meeting, Murati had been growing concerned that the board had not adequately prepared for the impact of Altman's dismissal, putting OpenAI at risk as a result. At the meeting, he also asked the board to provide a 30-minute deadline to explain the reasons for Altman's dismissal.

However, the board of directors decided that it should not be revealed that it was Murati who provided evidence showing Altman's management failures, and did not provide a clear explanation. The board tried to find a new CEO while Murati appeased the employees, but instead Murati led his colleagues in a rebellion against the board. Also, among the Altman faction, a story began to spread that the whole thing was a coup by Satkiver, and that Toner had backed it up. Satkiver, who had expected to get support from employees, was surprised by this reaction.

As a user of the social news site Hacker News commented, 'I'm more confused than before I read the article,' and 'This is nonsense. If Murati himself demanded an explanation, why didn't the board explain their decision?' These events suggest that there was confusion between Murati, Satskivar, and the board, and that they were not working well together.