OpenAI’s 2023 Leadership Crisis and What Happened Next

The events around OpenAI’s leadership have been extraordinary. In this November 20, 2023 video, I review the organization’s history, Sam Altman’s dismissal and what appeared to be the next move. The story was not actually over: Altman subsequently returned to OpenAI as CEO, so the video’s account is a snapshot of a fast-changing situation.

Berman begins with OpenAI’s 2015 founding as a nonprofit with a mission to advance AI for broad human benefit. He names founders including Sam Altman, Elon Musk, Ilya Sutskever and Greg Brockman and describes early research and experimentation. His account is a simplified history, not a complete list of founding contributors.

He highlights OpenAI Five, a system of five agents trained to play Dota 2, and the 2018 work on generative pre-training. The latter built on the Transformer architecture introduced in earlier research by Google-affiliated authors, while applying pre-training to language understanding.

In 2019 OpenAI created a capped-profit structure controlled by its nonprofit. Berman explains the stated objective of raising capital while retaining the mission and describes Microsoft’s investment. This describes the arrangement at the time of the video; the organization’s structure has changed since.

Berman shows a diagram of the 2023 corporate structure, in which the nonprofit board controlled entities associated with the for-profit operation. He emphasizes the board’s governance role and the mission’s definition of AGI, arguing that the arrangement created potential tension between commercial development and nonprofit oversight.

He moves through OpenAI’s product history, including GPT-3, Codex, DALL·E, ChatGPT and GPT-4. ChatGPT launched in November 2022; Berman’s reference to December and GPT-3.5 is a shorthand, not the exact launch date. He also recalls the November 2023 DevDay announcements of custom GPTs and a planned store.

On November 17, 2023, OpenAI announced that its board had removed Altman as CEO and named chief technology officer Mira Murati interim CEO. The board said it had concluded that Altman was not consistently candid in communications with it, impairing its ability to exercise its responsibilities. That was the board’s stated rationale, not a publicly detailed finding of particular misconduct.

The board also removed Greg Brockman as chair while initially expecting him to stay at the company; he resigned. Berman describes employees and technology leaders publicly supporting Altman, including many employees sharing expressions of solidarity. His favorable impression of Altman is personal commentary.

Berman reads from an employee letter challenging the board’s handling of Altman and Brockman. The letter alleged that the board had not supplied written evidence supporting its concerns and questioned its competence. These are claims made by the employees, not independently established findings.

The letter also objected to replacing Murati as interim CEO and said signatories might resign and join a newly announced Microsoft team. Berman describes attempts to negotiate a resolution and the display of employee solidarity through the phrase that OpenAI is nothing without its people.

Berman discusses a photograph of Altman visiting OpenAI with a guest badge during the negotiations. He relays rumors that a dispute over how to characterize Altman’s conduct contributed to an impasse. The video does not establish those private negotiations as fact.

At the time, the detailed reasons for the board’s action were not publicly clear. Berman considers a theory that a technical breakthrough and disagreement over the speed of releasing AI divided leadership and the board. He presents this as speculation; it should not be treated as an established explanation.

Later context: in March 2024, OpenAI published a summary of a WilmerHale review commissioned by a board committee. The summary said the prior board’s decision followed a breakdown of trust with Altman, and did not arise from concerns about product safety or security, development pace, finances or statements to partners. It also said the board acted within its discretion, while Altman’s conduct did not mandate removal.

Microsoft announced plans for Altman and Brockman to lead a new advanced AI research team. Berman speculates about potential financing and Altman’s future at Microsoft; those ideas were not confirmed. The planned Microsoft move was overtaken by subsequent negotiations and Altman’s return to OpenAI.

OpenAI appointed former Twitch CEO Emmett Shear as interim CEO after Murati. Berman reviews Shear’s stated plan to investigate the process, consult stakeholders and reform management. Shear said the board had not removed Altman over a specific disagreement on safety, although he did not publicly supply a complete explanation.

Berman reads Microsoft CEO Satya Nadella’s statement expressing continued commitment to OpenAI and announcing the planned research team. He summarizes what appeared on November 20 to be a new Microsoft destination and a change in OpenAI leadership, while anticipating further board and structural changes.

Ilya Sutskever publicly expressed regret for participating in the board’s actions and said he wanted to reunite the company. Berman interprets the message as a sign of reconciliation, while saying he still wants to know what happened internally.

The later outcome differed from the video’s apparent conclusion. On November 29, OpenAI announced Altman’s return as CEO, Brockman’s return as president, Murati’s return as CTO and a new initial board. The March 2024 review summary supplied additional context, though it did not publish every underlying detail of the dispute.