Two people testified under oath about what OpenAI was founded to be. Their accounts are incompatible. One of them is currently running OpenAI. [1]

Elon Musk's testimony characterized the founding agreement as unambiguous: OpenAI was established as a nonprofit specifically to prevent the concentration of AI capabilities in a for-profit entity. The mission was safety; the structure was the mechanism. When Sam Altman pursued the conversion toward commercial operations, Musk testified, he violated a founding commitment that Musk understood to be binding.

Altman's account described a founding process more fluid than Musk's framing allows — a set of evolving ideas among founders who shared concerns about AI safety but did not agree on a single institutional form. The for-profit structure, in Altman's telling, was not a betrayal of the founding but a necessary evolution of it. [2]

Both men are speaking under penalty of perjury. Both accounts cannot be fully accurate. The court has not yet adjudicated which is better supported by the contemporaneous documents — the incorporation papers, the early email chains, the founding agreements that both sides have introduced as exhibits.

Anna Weber's interest is not in who wins the lawsuit. It is in what the incompatibility reveals about institutions. OpenAI is one of the most consequential organizations in the world. Its governance — what it was built for, what it owes to whom, who has standing to hold it accountable — is being determined by a proceeding in which the two people who know the most about its founding cannot agree on what it was founded to do. That is not a legal problem. It is a legitimacy problem.

Institutions mean what their records say they mean, not what their founders remember years later. The documents will decide this. What they say about the founding purpose of AI's most powerful laboratory is not a narrow legal question.

-- ANNA WEBER, Berlin