Mira Murati, OpenAI's former Chief Technology Officer, testified under oath that CEO Sam Altman misled her about safety protocols for a new AI model and undermined her authority by sowing distrust among executives, in testimony presented during Elon Musk's lawsuit against OpenAI that goes beyond founder drama to reveal how the company's governance model is cracking under the weight of its trillion-dollar ambitions.
The deposition details are stark. In a video shown during the Musk v. Altman trial in federal court in Oakland, California, Murati said Altman inaccurately told her that OpenAI's legal team had determined a new model did not require review by the company's deployment safety board. When asked if Altman was being truthful, she replied, "No." She explained that she later consulted Jason Kwon, OpenAI's general counsel and current chief strategy officer, and found a "misalignment" between Kwon's position and Altman's. To ensure safety, Murati made sure the model underwent board review anyway. This was not a one-off disagreement. Murati said Altman complicated her work throughout her tenure by communicating different messages to different people, pitting executives against each other, and undermining her ability to do her job.
The testimony lands in the context of Musk's lawsuit, which alleges that Altman and OpenAI President Greg Brockman betrayed the company's founding agreement by transforming a nonprofit research lab into a for-profit entity that prioritises Microsoft over the public good. Musk, who co-founded OpenAI in 2015 and left in 2018, seeks $134 billion in damages and wants the company to revert to its nonprofit structure. Murati's statements bolster Musk's narrative by showing internal dysfunction at the executive level. She concurred with prior board concerns that Altman was not consistently transparent, which was the stated reason for his brief firing in November 2023. Murati herself was appointed interim CEO during that episode but criticised the board's handling, saying OpenAI was at "catastrophic risk of disintegration." She left OpenAI in 2024 to co-found her own AI lab, Thinking Machines.
That history is not incidental. OpenAI's governance has been a live issue since the 2023 board revolt, when the nonprofit board ousted Altman citing a lack of candor, only to reverse course under employee pressure and Microsoft influence. The company's hybrid structure, with a nonprofit controlling a capped-profit subsidiary, was designed to balance mission and incentives. But as OpenAI has scaled into a capital-intensive platform requiring billions in compute and revenue in the tens of billions, the tensions have become acute. Murati's testimony suggests those tensions played out in real time at the executive level, with Altman managing by creating divisions rather than building consensus around safety and deployment decisions.
For San Francisco founders, the dispute is a masterclass in founder control risk under extreme market pressure. OpenAI started as a small research lab with a clear mission to develop safe AGI for humanity. It has become a commercial juggernaut with trillion-dollar ambitions, Microsoft as its largest backer, and competitors circling. The governance model that worked at lab scale has struggled to scale with the company. A nonprofit board meant to oversee mission integrity proved unable to manage executive performance issues or strategic pivots. The for-profit arm became the dominant force, with the nonprofit relegated to symbolic oversight. Murati's account shows how those structural misalignments created day-to-day chaos for executives trying to ship products responsibly.
The mission promise is the part of the story that resonates most broadly. OpenAI's founding documents committed the company to open research and broad benefit, with a capped return to investors. Musk alleges that Altman and Brockman looted the nonprofit by building a closed, profit-maximising entity that functions as a Microsoft subsidiary. Whether that claim holds up in court, the testimony from Murati and others like former board member Helen Toner and co-founder Ilya Sutskever paints a picture of a company where personal trust eroded as commercial pressures mounted. The lesson for startups is clear: governance structures that feel adequate in the early days can become liabilities when the company scales into a category-defining platform. Founders who promise mission alignment need to design boards, incentives, and decision processes that can withstand the pressure of trillion-dollar valuations and global scrutiny.
OpenAI's structure may not survive its compute ambitions intact. The nonprofit model made sense when the company was burning through $100 million a year on research. It looks increasingly unworkable when annual compute spend alone approaches $10 billion and revenue targets are in the tens of billions. A fully for-profit conversion may be inevitable, but it would require rewriting the founding agreement and facing the political and reputational cost. Murati's testimony is a reminder that the internal cost of maintaining the hybrid model is already high, and the external legal cost is now material too. The company that defined the AI era may need to redefine its own governance to survive it.
Also read: Ethos raises $22.75M from a16z to rebuild expert networks around voice onboarding • Alphabet and Amazon's AI startup gains are now driving more than half their headline earnings • Match Group is slowing hiring to fund AI and consumer internet has its new margin script
