The trial meant to expose OpenAI's nonprofit betrayal is instead becoming a deposition on whether the most powerful CEO in AI can actually manage a company.
The Summary
- Text messages between Sam Altman and Mira Murati during his 2023 ouster show him pleading "Still don't want me?" while she responds "They don't want you" — painting a picture of desperation, not control
- Multiple former executives testified that Altman "told people what they wanted to hear" and created "chaos" through poor communication, criticisms that predate his brief firing
- The Musk v. Altman trial, ostensibly about whether OpenAI abandoned its nonprofit mission, is revealing the management dysfunction behind the world's most valuable AI company
- Shivon Zilis, former OpenAI board member and mother of four of Musk's children, testified live about Altman's leadership failures
The Signal
The November 2023 OpenAI board crisis looked like a coup from the outside. From the inside, the text messages reveal something messier: a CEO locked out of his own building, texting his interim replacement for updates. When Altman asked Murati how things were going "directionally," she answered "very bad." When he pressed whether the board wanted him fired or something else, she clarified: "Yes, for you to be gone."
This wasn't strategic maneuvering. This was a man scrambling in real time while his employees decided whether to keep him. The texts show Altman asking to be "officially invited" to the office, suggesting he'd been physically barred. Within 48 hours, he'd be reinstated. But the exchange captures the reality underneath the triumphant return narrative: for two days, the CEO of the most important AI company in the world had no idea if he still had a job.
"They're convinced about their decision."
The trial testimony goes deeper than one bad weekend. Mira Murati told the jury that Altman "told people what they wanted to hear," a polite way of describing someone who makes conflicting promises. Former board member Shivon Zilis testified live about management problems that predated the firing. These weren't post-hoc rationalizations. They were documented complaints from people running the company.
Altman himself has acknowledged the problem. In a recent blog post responding to a New Yorker piece questioning his honesty, he wrote that he's "not proud of being conflict-averse," which caused "great pain" and created a "huge mess" for OpenAI. That's an interesting admission for someone leading a company racing to build AGI. Conflict aversion in a CEO usually means unresolved tensions, unclear priorities, and teams working at cross purposes because no one wants to force a decision.
Key tensions that emerged from testimony:
- Altman's tendency to tell different stakeholders what they wanted to hear
- Communication breakdowns described as creating "chaos"
- A pattern of behavior that preceded the dramatic November 2023 firing
The irony is that this trial was supposed to be about Musk's claim that OpenAI betrayed its nonprofit mission. Instead, it's becoming a public autopsy of how OpenAI actually operates. The Musk lawsuit seeks over $100 billion, alleging the founders "stole" a nonprofit. But the testimony is revealing something potentially more damaging: that the company building the technology that might reshape human work and intelligence has had serious internal dysfunction at the leadership level.
The Implication
If you're building in the agent economy, pay attention to this dynamic. The companies that will define Web4 aren't just competing on model performance. They're competing on execution, and execution requires management that can make hard calls and communicate clearly. Altman's conflict aversion worked fine when OpenAI was a research lab. It becomes a serious liability when you're productizing AGI and every decision has billion-dollar stakes.
For anyone trying to figure out which AI company to build on, bet with, or work for: management competence is now part of the technical risk assessment. A CEO who can't manage conflict will struggle to ship agents that work reliably at scale. Watch how leadership teams handle disagreement. That's the real signal.