Sam Altman got fired and then unfired in 5 days

On Friday, November 17th, OpenAI’s board of directors fired Sam Altman. By the following Wednesday, he was back as CEO with a new board.

Five days. The most chaotic five days in the history of AI.

What happened

The board, which included AI safety researchers and nonprofit governance leaders, told Altman he was fired for not being “consistently candid” with the board. They didn’t provide specific examples. They didn’t give warning. They called a meeting, told him, and announced it publicly.

What followed was a corporate drama that played out in real time on Twitter and in The Information and the New York Times:

Day 1 (Friday): Altman fired. President Greg Brockman quits in solidarity. Interim CEO named.

Day 2 (Saturday): Investors scramble. Microsoft, OpenAI’s largest investor, reportedly blindsided. Rumors of Altman joining Microsoft directly.

Day 3 (Sunday): 700 of OpenAI’s 770 employees sign a letter threatening to resign unless the board steps down and Altman is reinstated.

Day 4 (Monday): Negotiations. The board considers reinstating Altman. Talks break down. Resume. Break down again.

Day 5 (Tuesday/Wednesday): Altman returns as CEO. The board is restructured. The safety-focused members are out. Business-oriented members are in.

What it revealed

The drama revealed something important beneath the corporate chaos: the people building the most powerful AI systems in the world don’t agree on how to govern them.

OpenAI was founded as a nonprofit with a mission to build safe AGI for the benefit of humanity. It created a capped-profit subsidiary to raise money. The board’s job was to ensure the mission took precedence over profit.

The board fired the CEO, apparently over concerns about the pace of commercialization versus safety. The employees, who have equity in the capped-profit subsidiary, revolted. Microsoft, which has invested $13 billion, applied pressure. The money won. The safety-focused board members were replaced.

I’m not saying the board was right to fire Altman. The process was botched. The lack of transparency was damaging. The execution was terrible.

But the underlying tension, between “build fast and capture the market” and “slow down and make sure this is safe,” is real. And the resolution of that tension, in favor of speed, at the most important AI company in the world, should concern all of us.

The governance question

Who should govern AI development? The people building it (who understand it but have financial incentives to build faster)? The investors (who understand money but not safety)? Governments (who understand regulation but move slowly)? Independent boards (who might be thoughtful but lack leverage)?

OpenAI tried the independent board model. It lasted until the board tried to exercise its independence. Then the employees and investors overruled it.

The experiment failed. Not because the model was wrong in principle, but because a nonprofit governance structure can’t survive when the economic forces it’s meant to constrain are worth billions of dollars.

What I keep thinking

The CEO of the most important AI company was fired and rehired in five days. The safety-oriented board members were replaced by business-oriented ones. The message, whether intended or not, is: AI development will be governed by the people who profit from it, not the people who worry about it.

I don’t know what the right governance model is. I don’t think anyone does. But I know that the current model, where the internal drama of one company determines the trajectory of the most powerful technology in history, is not it.

Five days. That’s how long it took to resolve the question of who governs AI at OpenAI. The answer was: money.

I wish I were surprised.

Related thinking: