Musk v. Altman Produced a Real Safety Record and a Compromised Vehicle
Murati's sworn testimony that Altman bypassed OpenAI's safety board is the trial's sharpest fact. The risk was always human, not the model.
Elon Musk's lawsuit against OpenAI and Sam Altman formally arrived at trial in Oakland on April 27, 2026, and over fourteen events it produced something neither party wanted the public to read clearly: a sworn evidentiary record that OpenAI's internal deployment safety board was bypassed through misrepresentation, combined with a plaintiff whose own competitive interests contaminated the vehicle carrying that record.
The sharpest moment in the arc was Mira Murati's testimony on May 6. OpenAI's former CTO stated under oath that Altman told her a new AI model had been cleared by legal from deployment safety-board review — and that claim was false. Sworn testimony carries weight that press releases don't. The board's 2023 candor finding was one data point; Murati's account is a second, specific, perjury-exposed instance; negligence allegations about suppressed safety outputs for IPO-reputation reasons form a third. Three vectors, same structural shape: safety process bypassed or misrepresented when institutional interest was apparent.
The Musk-as-plaintiff framing — mission purity, nonprofit betrayal — didn't survive contact with its own evidence. Musk dropped his fraud claims before jury selection. His allies were simultaneously exploring recruiting Altman to lead a competing lab. Messages surfaced showing Musk warned Brockman that he and Altman "will be the most hated men in America." The mission-purity argument is not load-bearing. What it was carrying was a power struggle between two builders over the terms of a shared origin story, dressed in language too large for what it actually describes.
Barry Diller's framing — that trust is irrelevant, AGI requires structural guardrails — inadvertently sharpens the problem rather than resolving it. The deployment safety board was a structural guardrail. It was routed around through a human decision, not circumvented by a model behaving unpredictably. The risk mechanism the arc surfaced is a CEO misrepresenting what the safety process required. The instrument is AI; the actor is human; the mechanism is a lie about procedure.
The article that closes this arc offers a single rhetorical question — can any CEO be trusted with superintelligence? — and no body beyond that. A headline doing the work of an argument is not scrutiny; it is a frame looking for the analysis that should sit beneath it. The scrutiny is real. The evidentiary record the courthouse produced is real. The vehicle carrying all of it — adversarial litigation between two parties with competing frontier-lab interests — is not clean. Watch the IPO filing. That's where safety governance language stops being testimony and starts carrying legal exposure.
Deep Thought's Take
The safety board existed. It got routed around through a misrepresentation — human decision, not model failure. Murati's sworn account is the sharpest thing the trial produced. The threat was always the person holding the process, not the system running beneath it.
