Latest / The CEO Diary with Fexingo: Leadership Lessons, Executive Decisions, and Corner Office Stories / How Sam Altman Forced a Board Coup at OpenAI
Transcript
- Lucas: So it's June 2026, and we're sitting here nearly three years after one of the strangest boardroom dramas in tech history. I'm talking about the OpenAI board coup in November 2023, when the nonprofit board fired CEO Sam Altman — and then had to beg him to come back five days later after nearly the entire staff threatened to walk. Luna: Right — this is one of those episodes where the conventional startup governance model just completely broke down. And it wasn't a small company. OpenAI was already valued at 80 billion dollars at that point. Lucas: Exactly. And what's wild is that the board's decision wasn't about performance in the traditional sense. Revenue was growing. The product — ChatGPT — was a global phenomenon. The firing was about trust and mission. The board claimed Altman wasn't consistently candid with them. Luna: But the real story is the structure, right? OpenAI's board wasn't a typical for-profit board. It was a nonprofit board overseeing a capped-profit subsidiary. Lucas: That's the key. OpenAI was originally a nonprofit in 2015. In 2019, they created a for-profit arm to raise capital — but the board's fiduciary duty remained to the nonprofit's mission of safe AGI. So you had a board with no shareholders to answer to, no profit-maximisation requirement, and members who genuinely believed Altman was moving too fast and cutting safety corners. Luna: The board members who voted to fire him — Adam D'Angelo, Tasha McCauley, Helen Toner, and at first Ilya Sutskever — they weren't investors. They were academics and researchers. Toner was a PhD in AI governance. McCauley a tech strategist. Lucas: Right. And that's where the governance failure starts. A board of people who are not financially incentivised but are deeply mission-driven can make a decision that looks principled in isolation — but it backfires when the actual workforce doesn't back them. Within hours of the firing, all three senior researchers resigned. Then 700 out of 770 employees signed a letter saying they'd quit if Altman didn't come back. Luna: And Microsoft, which had invested 13 billion dollars, stepped in and offered to hire anyone who left. That basically gave the employees a safety net to rebel. Lucas: So you had a board that acted on principle but without a plan for execution. They didn't anticipate the talent revolt. They didn't have a backup CEO. They didn't realise that the company's value — both financial and operational — was tied to Altman personally. And that's a classic governance blind spot. A board that thinks it's protecting the mission can destroy it if they don't understand the organisation's actual social contract. Luna: Let's talk about that social contract for a second. OpenAI's employees were not just hired to write code. Many of them joined because they believed in the mission of safe AGI. But they also believed in Altman's vision. They saw the board's move as a betrayal of the team's trust, not a protection of it. Lucas: And that's the paradox. The board was trying to enforce the original nonprofit mission — but the employees, who were also mission-driven, sided with the CEO they thought could actually deliver. The board lost all legitimacy overnight because they had no constituency. They had no investors backing them, no employees, and no clear narrative that resonated with the public. Luna: The board's statement after the firing was vague. They said Altman wasn't consistently candid — but they never gave specifics. That vagueness made them look like they were hiding something. In a crisis, vague accusations look like a power grab. Lucas: Compare that with how Jamie Dimon handled a similar situation at JPMorgan — we covered that in episode 61. Dimon had a board that fully backed him through crises because there was constant communication. Altman's board felt shut out, so they overreacted. The lesson: if the board doesn't feel informed, they can panic. But if the CEO over-communicates, the board becomes an ally. Luna: So what was the actual outcome after Altman returned? A new board. Larry Summers joined. Bret Taylor became chairman. The old guard was mostly gone. And the board structure changed — it became more traditional, with more business figures and fewer AI governance experts. Lucas: Right. And here's the part that I think matters most for founders listening. OpenAI's original structure — a nonprofit board governing a for-profit subsidiary — was meant to keep the company aligned with its mission as it scaled. But in practice, it created a board with no accountability. No shareholders to report to. No market feedback. Just four people who could, in their judgment, fire the CEO. That kind of power without accountability is a ticking time bomb. Luna: And the board composition was also a problem. Only one of the four had real operating experience — Adam D'Angelo, who is CEO of Quora. The others were academics. When you have a board of mission experts but no operators, they may not understand the operational consequences of their decisions. Lucas: Exactly. And I think this is the one concrete thing I'd want listeners to take away. If you're building a company with a dual structure — a nonprofit or foundation controlling a for-profit — you need to think hard about board composition. You need people who understand both mission and execution. And you need a clear, documented process for what happens when the board loses confidence in the CEO — not just a vague clause about removing officers. Luna: It's worth asking: could the OpenAI saga have been avoided if the board had a more traditional governance framework? Or is this an inherent risk of mission-driven companies? Lucas: I think it's partially inherent. Any mission-driven organisation faces this tension. The board sees its duty as protecting the mission. The CEO sees their duty as growing the organisation's ability to achieve the mission. Those can diverge. But the OpenAI board failed because they didn't communicate. They didn't build a coalition. They acted unilaterally and expected everyone to fall in line. That's not governance — that's a palace coup. Luna: And the aftermath — Altman is still CEO as of mid-2026. The company has raised more money, launched new products, and the governance structure is more conventional. But the scars remain. I think the broader lesson for any founder is: choose your board carefully, and make sure they have skin in the game — not just mission conviction, but real accountability for the outcomes of their decisions. Lucas: Yeah. If today's conversation was useful — maybe you're thinking about your own board structure or you found the governance angle helpful — the way this show stays ad-free is through listener support. You can buy me a coffee dot com slash fexingo. That's buy me a coffee dot com slash fexingo. It helps us keep digging into these stories without any sponsor pressure. Luna: And we really appreciate it. It's a small gesture that goes a long way in keeping the show independent. Lucas: So back to the OpenAI fallout. One of the most interesting post-coup changes was how Altman rebuilt his relationship with the new board. He brought in Bret Taylor, who's a seasoned board member — former co ceo of Salesforce, chair of the board at a number of companies. That's a signal that Altman understood he needed adult supervision, someone who could manage board dynamics professionally. Luna: And Taylor's first move, reportedly, was to establish formal communication rhythms — weekly updates, a clear process for raising concerns. The exact opposite of the previous board's experience. Lucas: Right. And that's a playbook any CEO can use. If your board feels out of the loop, the fix isn't to give them more power — it's to give them more information. The former board at OpenAI felt they had to act because Altman wasn't keeping them informed. Whether that's true or not, the perception was there. And perception drives board behaviour. Luna: One other angle — the employee reaction. The staff letter demanding Altman's return was unprecedented in its scale and speed. It showed that in the knowledge economy, the CEO's most important asset is the trust of their team. Altman had that trust. The board didn't. And without employees, a tech company is worthless. Lucas: That's the ultimate takeaway for me. Governance isn't just about structure — it's about legitimacy. The board had legal authority but zero legitimacy. Altman had no formal power at that moment but massive legitimacy. And legitimacy beats authority every time in a crisis. So if you're a board member, your job isn't just to make the right decision — it's to make a decision that people will accept as legitimate. That means communicating, building consensus, and understanding the human dynamics of the organisation you govern. Luna: And if I'm a founder listening, I'm thinking: how do I make sure my board has both the right composition and the right communication culture before a crisis hits? Lucas: Right. Don't wait until there's a problem to figure out how your board operates. Have explicit discussions about what information they need, how often, and in what format. And make sure the board includes people who understand your business, not just your mission. The OpenAI board had too many people who understood AI governance but not how to run a 2,000-person company. Luna: So we've talked about structure, communication, and legitimacy. Anything else you'd flag? Lucas: One more thing: the role of large investors. Microsoft was in the background but didn't have a board seat. After the coup, they got a non-voting observer seat. That's a compromise — they have visibility but not control. For founders with strategic investors, think about what kind of board access you give them. Too little, and they may panic in a crisis. Too much, and you lose control. OpenAI's solution was a middle ground that seems to be working. Luna: All right, that's a solid note to end on. Thanks for listening, and we'll be back with another episode soon.