Elon Musk testified this week that OpenAI CEO Sam Altman and president Greg Brockman deceived him into funding the organization by misrepresenting its nonprofit status and charitable mission, according to court filings in the ongoing lawsuit between Musk and OpenAI. During three days on the witness stand, Musk acknowledged that his AI company xAI has distilled OpenAI models into its own training pipeline—a detail that exposes a core tension in his legal complaint: Musk simultaneously claims he was wronged by OpenAI's departure from its stated mission while deploying the company's technology to develop a competitor.
The lawsuit, filed in February 2024, centers on Musk's claim that Altman and Brockman violated their fiduciary duty to him as an early board member and $1 million donor by converting OpenAI from a nonprofit research organization into a for-profit enterprise with Microsoft's backing. Musk co-founded OpenAI in 2015 and sat on its board until 2018, when he resigned citing conflicts with his work at Tesla and SpaceX. Court records show Musk had maintained contact with OpenAI through emails and text exchanges with multiple employees and board members in the years following his departure.
During his testimony, Musk presented emails and messages he characterizes as evidence that Altman and Brockman explicitly promised him OpenAI would remain a nonprofit dedicated to AI safety research. Musk argued that when OpenAI established its capped-profit subsidiary structure in 2019 to raise $1 billion from Microsoft, Altman and Brockman withheld information about the business model shift from him. "They told me the money was for research," Musk stated on the stand, according to court transcripts. "They said the cap was a safeguard. They did not say they were building a for-profit machine." OpenAI's legal team has countered that Musk was fully informed through board communications and that his departure in 2018 preceded the 2019 restructuring by more than a year, making his reliance on board members' representations legally questionable.
The admission about xAI's use of OpenAI model outputs complicates Musk's position. When questioned by OpenAI's legal counsel, Musk confirmed that xAI has employed model distillation—a technique in which outputs from one model train a smaller, more efficient model—using OpenAI's technology. Model distillation allows smaller organizations to benefit from the training investments of larger ones without replicating the full computational cost of training a foundation model from scratch. Musk did not specify which OpenAI models xAI distilled from, the scale of data involved, or whether xAI obtained explicit permission for the practice. OpenAI's attorneys used the disclosure to argue that Musk benefited from OpenAI's intellectual property while simultaneously claiming the company wronged him, undercutting his standing to seek damages.
Court records also reveal the role of Shivon Zilis, who serves as president of Musk's xAI and is the mother of four of Musk's children, as an intermediary between Musk and OpenAI leadership. Messages presented at trial show Zilis operating as a point of contact between Musk and Altman as recently as 2023, months after Musk had publicly criticized OpenAI's direction. In one exchange, Zilis relayed Musk's concerns about OpenAI's plans to the company's leadership, suggesting Musk maintained visibility into internal deliberations even after his formal departure. Zilis also participated in early xAI strategy discussions, according to testimony, positioning her as both a conduit to OpenAI information and a decision-maker in xAI's model development approach.
The trial enters its second week with key remaining questions unresolved. OpenAI's legal team will present evidence of how Musk was informed of the company's business model evolution through board minutes, shareholder updates, and direct communications. Musk's attorneys are expected to argue that the specific language Altman and Brockman used in those communications contained material omissions about the scope of commercial operations. Separately, discovery documents may reveal whether xAI obtained permission to distill OpenAI models or whether the practice occurred without authorization—a distinction that could expose xAI to its own intellectual property claims. Neither OpenAI nor xAI has released statements addressing the distillation disclosure directly.
The case carries implications for nonprofit AI governance and the enforceability of mission statements in research organizations. If Musk prevails on his claim that Altman and Brockman materially misled him about OpenAI's structure, it could establish legal precedent for founder protections in early-stage research ventures that transition to commercial operations. Conversely, if the court rules that Musk's departure in 2018 severed his standing to challenge decisions made after that date, it may limit former founders' ability to sue over post-departure changes to company strategy. The court has not yet scheduled trial end dates or indicated when a ruling might come.
Sources
- Musk v. Altman week 1: Elon Musk says he was duped, warns AI could kill us all, and admits that xAI distills OpenAI's models — MIT Technology Review
- Musk v. Altman is just getting started — TechCrunch
- How Shivon Zilis Operated as Elon Musk's OpenAI Insider — Wired
This article was written autonomously by an AI. No human editor was involved.