The board that fired him in 2023 can't do it again without a supermajority, and the timing of that rule change says everything about who really runs OpenAI now.
The Summary
- OpenAI quietly changed its bylaws in October 2025 to require a two-thirds supermajority of non-employee directors to fire the CEO, up from a simple majority under the old nonprofit structure.
- Ilya Sutskever testified in the Musk v. Altman trial, defending his role in the 2023 ouster attempt despite being estranged from OpenAI.
- Altman now needs only one-third board support to stay in power, inverting the accountability structure that existed when the nonprofit board attempted to remove him.
- The governance change came during OpenAI's for-profit transition, effectively locking in Altman's position while the company restructures around commercial interests.
The Signal
Sam Altman got fired in November 2023. By October 2025, OpenAI had rewritten the rules to make sure it couldn't happen again. The new bylaws, analyzed by Columbia law professor David M. Schizer as part of Elon Musk's lawsuit against OpenAI, require a two-thirds supermajority of non-employee board directors to remove the CEO. Under the old nonprofit structure, a simple majority was enough.
Do the math. Altman now needs only one-third of the board on his side to keep his job. The accountability mechanism flipped from "most directors want you gone" to "most directors can't make you leave."
"The governance change inverts power from the board to the CEO, requiring consensus to remove rather than consensus to retain."
This isn't how normal companies work. The supermajority structure contrasts sharply with most large US public companies, which typically operate on simple majority votes for CEO removal. The change came during OpenAI's for-profit conversion, the same transition that shifted control from the nonprofit mission-focused board to a structure optimized for commercial success and investor returns.
The timing matters. Schizer's analysis, conducted in November 2025 and released in court documents this week, shows the bylaw change happened right as OpenAI was cementing its transformation from nonprofit research lab to $150+ billion for-profit entity. When you're restructuring a company around a CEO's vision and fundraising power, you don't want the board second-guessing that CEO every quarter.
Meanwhile, Ilya Sutskever took the stand Monday to defend his role in the original firing attempt. The former chief scientist, now estranged from OpenAI, testified in the Musk v. Altman trial that he acted to prevent the company from being "destroyed." He's defending a decision that the new bylaws are explicitly designed to prevent from ever happening again.
Key tensions now visible:
- The board that tried to fire Altman had majority support but lost the power struggle
- The new board structure prevents that scenario from repeating
- Sutskever defends the original ouster while the company locks in rules against future attempts
The for-profit conversion gave OpenAI legal cover to rewrite governance from scratch. The supermajority rule wasn't an accident. It was the lesson learned from November 2023: never let a simple majority of directors tank the company over safety concerns, mission drift, or personality conflicts again.
The Implication
If you're building in the agent economy, watch how power consolidates around the people who can ship product and raise capital. OpenAI's governance shift is a template: move fast, convert to for-profit, lock in the CEO who can execute. Safety-focused boards are a liability when the race is measured in months, not years.
For everyone else, this is what Web4 looks like in practice. The companies building the agent infrastructure aren't governed by mission statements or nonprofit charters anymore. They're governed by whoever controls one-third plus one board seat. That's enough to keep building, keep shipping, and keep the revenue growing while everyone else argues about alignment.