Analysis
Sam Altman told Fortune that running Y Combinator under Paul Graham taught him to govern rather than rule, and that OpenAI "won't get every decision right." It is a modest framing from the executive of a company whose governance has been the single most scrutinized structure in technology since November 2023.
The context is unavoidable. Altman was removed by OpenAI's board and reinstated within five days in 2023; the nonprofit-to-PBC restructuring consumed most of 2025 and drew formal attention from the California and Delaware attorneys general; the preparedness team was dissolved in August; and a string of senior departures preceded every governance milestone. "Govern, not rule" is a claim that invites checking against that record rather than one that settles it.
The Y Combinator comparison is doing real work, though. YC's structure is genuinely federal: partners make independent investment calls, batch companies operate with near-total autonomy, and the president's leverage runs through norms and selection rather than direction. Altman ran that organization from 2014 to 2019 and grew it from a seed program into the most reliable pipeline in venture. If the analogy holds, it predicts an OpenAI that pushes decisions to product teams and holds a small number of lines centrally.
“Altman ran that organization from 2014 to 2019 and grew it from a seed program into the most reliable pipeline in venture.”
The practical test arrived the same week. California enacted a chatbot-safety law that OpenAI negotiated directly and, per Fortune's reporting on those talks, the company traded specific concessions to shape a statute that other states will copy. That is governing in the political sense -- accepting binding constraints in exchange for influence over their content -- and it is a meaningfully different posture from the industry's 2023 lobbying stance.
The skeptical read is that governance language tends to arrive on schedule before a listing. Altman said last week that OpenAI is delaying its IPO on safety grounds, and Pulse has tracked the OpenAI timeline slipping repeatedly through the year. Investors buying into a company with a capped-profit history and an unusual board will want the governance story told confidently, and it is being told confidently.
What would make it credible is structural rather than rhetorical: an independent safety board with the documented authority to block a release, published minutes, and a named person who is not the CEO holding that authority. Absent that, "we won't get every decision right" is an apology issued in advance.
