Murati Testifies Altman Bypassed OpenAI Safety Board Under Oath
Three independent source batches converged on the same courtroom disclosure on 9 May. Mira Murati, former OpenAI CTO and brief interim CEO during Sam Altman's November 2023 firing, testified under oath that Altman lied to her about safety review clearance for a new AI model — producing the most specific named-source disclosure yet in Musk's nonprofit-breach lawsuit against OpenAI.
What the Source Actually Says
Murati testified that Altman told her OpenAI's legal department had determined a new AI model did not need to go through the company's deployment safety board. She then went directly to general counsel Jason Quan — and found the accounts did not match. "I confirmed that what Jason was saying and what Sam was saying were not the same thing," she testified. Rather than take Altman's word, she routed the model through the safety board anyway.
Separately, Greg Brockman's personal journal has emerged as a key exhibit. Brockman says he stopped writing about OpenAI in it in 2023; it has nonetheless entered as star evidence, adding a contemporaneous written record to what would otherwise be disputed oral accounts. Brockman's personal OpenAI stake was also disclosed at $30 billion during proceedings — directly relevant to Musk's claim that the nonprofit structure was used to enrich founders.
The court released the full text-message exchange between Murati and Altman from the weekend Altman was fired. The messages show Altman probing whether reinstatement was possible while Murati relayed the board's position in real time. One message captures Altman's reaction on learning who the replacement CEO would be: "New guy is rando Twitch guy." Satya Nadella was eventually brought into the thread as a stabilising force. Gary Marcus — who testified alongside Altman at the US Senate in 2023 — flagged Altman's upcoming cross-examination as the highest-stakes remaining moment, noting Altman is "one of the world's most convincing (though not always truthful) talkers" who is rarely subjected to genuinely hard questioning.
Strategic Take
Murati's testimony names Jason Quan by name and describes a documented verification step — specificity that is hard to strategically reframe. Combined with the Brockman journal and the firing-weekend text record, the trial is building a hard factual picture of AI safety governance at the most influential lab in the field. Altman's cross-examination is the next key moment.
