The Verge discloses only an RSS snippet here, not Mira Murati’s deposition text, exhibit pages, or question-by-question context. Thin source material, but the underlying issue is not thin: OpenAI’s November 2023 Altman firing has moved from boardroom mystery into searchable litigation evidence in Musk v. Altman.
My read is simple: don’t treat this as workplace gossip. OpenAI’s line that Sam Altman was “not consistently candid in his communications with the board” remains one of the most expensive vague sentences in AI. Expensive because it nearly cost OpenAI its CEO, staff cohesion, Microsoft confidence, and product cadence. Vague because the board never publicly explained which concrete acts triggered that claim. Murati’s deposition and trial exhibits now matter because litigation is forcing details into the open that OpenAI’s governance process never supplied.
The snippet does not disclose what Murati actually said. The title says her deposition pulled back the curtain; the body says much of the material centers on the former CTO. That is enough to explain why practitioners should care, but not enough to assign blame. Murati was CTO, then briefly interim CEO during the crisis. She sat close to the technical core and was temporarily placed at the center of executive power. Her testimony naturally touches three questions: what Altman told the board, what Murati knew at the time, and how employees and leadership moved back toward Altman within days. The RSS version gives none of those answers.
The useful frame is governance under model-lab pressure. OpenAI is not a normal SaaS company with a clean board-shareholder-management stack. It has a nonprofit parent, capped-profit history, Microsoft dependency, ChatGPT distribution, API revenue, enterprise contracts, and frontier-model safety claims all packed into one structure. The 2023 firing exposed the stress point: can a mission-oriented board actually constrain a CEO who controls staff loyalty, commercial momentum, and partner confidence? The lived answer was no. Altman returned within days, the board was remade, Bret Taylor and Larry Summers joined, and practical power shifted toward executives plus capital partners.
The comparison to Anthropic and Google DeepMind is useful. Anthropic made its Public Benefit Corporation and Long-Term Benefit Trust structure more explicit from the start, even if I still have doubts about how hard those constraints bite under heavy cloud and investor dependence. DeepMind sits inside Alphabet, where safety governance ultimately lives under a parent company with product and cloud incentives. OpenAI’s position is more awkward. It wants the moral legitimacy of a nonprofit mission while operating at ChatGPT and Azure speed. The 2023 board fight was the first public rupture in that arrangement.
I also don’t fully trust the “pulled back the curtain” framing until the full deposition is available. Trial materials are not neutral documentary footage. Musk v. Altman has a strategic purpose: Musk’s side wants to show mission drift; OpenAI’s side wants to narrow the dispute into governance and commercial claims. Depositions are sworn, but media excerpts still impose sequence, emphasis, and omission. Without the full transcript, “centered on Murati” is a flag, not a verdict.
Still, I’ll make one firm call: OpenAI is still paying interest on that 2023 announcement. A mature frontier-model company should not need lawsuit exhibits to clarify its most important governance rupture. After model autonomy, agent permissions, and deployment risk all intensified through 2024 and 2025, the serious question is not who betrayed whom during Thanksgiving week. It is who can pause a release, who can overrule the CEO, what evidence triggers removal, and what disclosure obligations exist when leadership withholds information.
Murati’s personal arc matters less than the institutional one. She later left OpenAI and founded Thinking Machines Lab, which already tells you the 2023 leadership cluster did not remain aligned. The sharper point is that OpenAI’s governance story still leans on trust where it needs auditability. If key facts keep surfacing through Musk litigation and selective media excerpts, OpenAI cannot credibly claim it has solved the power-control problem around frontier AI leadership.