Sam Altman gave OpenAI’s board and colleagues information that was simply wrong — false references to safety approvals, nonexistent legal opinions. That is the core finding of a New Yorker investigation published Wednesday, built on an 18-month reporting effort that pulled from internal documents, notes attributed to co-founder Ilya Sutskever, material from executive Dario Amodei, and more than 100 interviews.
The piece is not a hit job. It is a slow, methodical reconstruction of how a company originally chartered as a nonprofit veered hard into commercial territory, with its CEO telling people things that were not true along the way. The investigation describes “several instances” of Altman giving incorrect information. It does not say he lied. It says he gave incorrect information. The distinction matters, but only so much.
OpenAI’s original mission was safety-first, nonprofit, transparent. That mission eroded as development accelerated. Safety teams were reduced. The company now carries large financial commitments that some board members view as risky. The governance question — who watches the watcher — runs through the entire piece.
The New Yorker’s reporters had access to Sutskever’s notes. Sutskever was a co-founder and, for a time, the company’s chief scientist. He was also on the board that briefly fired Altman in November 2023, then reinstated him days later. Those notes, combined with material from Amodei — now CEO of rival Anthropic — and more than 100 other interviews, paint a picture of a CEO who operated with what the piece calls a loose relationship with factual accuracy when it suited his goals.
One example: Altman referenced safety approvals that did not exist. Another: he cited legal opinions that had not been written. The board was left to make decisions on information that turned out to be false. That is not a minor procedural glitch. That is a board being misled by its own CEO.
The shift from nonprofit to commercial structure accelerated under Altman’s leadership. The company originally capped returns for investors. That cap was removed. The company now sells access to its technology. It has a for-profit arm. The safety teams that once had real power to slow or stop deployments have been cut. The board’s composition changed. The original nonprofit board that could fire a CEO was replaced by a board that, in practice, cannot.
Altman himself is the central figure in the story. He has defended his leadership and the company’s approach to safety in earlier board discussions. The New Yorker piece is the most detailed public account yet of what was happening inside those discussions.
The implications are straightforward. If a CEO gives incorrect information to a board, the board cannot govern. It cannot make informed decisions. It cannot hold leadership accountable. The company’s structure — originally designed to prevent exactly this scenario — was dismantled over time. The safety teams that were supposed to act as internal checks were reduced. The financial commitments that now hang over the company were taken on by a leadership that, according to the investigation, was operating on faulty information.
OpenAI has previously defended its leadership and safety approach amid earlier board turmoil. The New Yorker investigation adds a new layer of evidence to that debate. It is detailed, sourced, and specific. It does not rely on anonymous accusations. It relies on documents, notes, and named sources.
The company now faces a reckoning not just with regulators or critics, but with its own founding principles. A nonprofit that shifted to commercial. A safety-first mission that accelerated development. A board that was given wrong information. A CEO who gave it.
