SAN FRANCISCO — A civil lawsuit pitting two of Silicon Valley’s most prominent figures against each other concluded its evidentiary phase in a San Francisco federal courtroom this week, and the testimony, documents, and competing narratives that emerged over eleven days painted an unusually candid portrait of how decisions at the frontier of artificial intelligence are actually made — and contested.
The case, brought by Nexus Ventures founder and technology executive Eliot Hartmann against Meridian AI, the nonprofit-turned-commercial AI research organization, centers on allegations that Meridian’s leadership improperly converted billions of dollars in charitable assets into private equity stakes when it restructured its governance in 2024. Meridian’s chief executive, Daniel Osei, is accused of orchestrating the conversion in ways that personally enriched early investors at the expense of the public interest mission the organization had publicly championed. Osei has denied all allegations and filed counterclaims alleging defamation and tortious interference.
Internal documents introduced by plaintiffs’ attorneys showed that Meridian’s board voted to approve the structural conversion in a meeting that lasted just 34 minutes and that three of the seven board members participated by phone while traveling internationally. Legal experts called the procedural record thin for a decision involving an estimated $18.7 billion in assets. The documents also showed that outside counsel had flagged potential fiduciary concerns in a memo circulated four days before the vote, a memo that several board members testified they had not read before casting their ballots.
Testimony from a former Meridian vice president of strategy revealed that the organization had maintained informal priority access arrangements with at least four commercial partners, giving those companies early access to AI model outputs in exchange for cloud computing credits. The witness said the arrangements were never disclosed to the full board and were described internally as research collaborations. Regulatory attorneys watching the trial said the arrangements could draw scrutiny from the Internal Revenue Service and state attorneys general offices in several jurisdictions.
Hartmann’s legal team introduced a series of private messages exchanged between Osei and two co-founders of a competing AI startup, in which Osei appeared to discuss strategies for limiting a rival researcher’s access to computing resources. The rival researcher, who was not a party to the lawsuit, had publicly criticized Meridian’s safety practices. Osei’s attorneys argued the messages were taken out of context and that no action was ever taken, but the exchange attracted significant attention in the courtroom and among technology journalists covering the proceedings.
Expert witnesses called by both sides offered sharply divergent valuations of Meridian’s core AI model assets at the time of the conversion — a gap of more than $6 billion between the two estimates. The valuation methodology used by Meridian’s own financial advisers, which relied heavily on proprietary projections that were not shared with regulators at the time of the restructuring, was challenged extensively in cross-examination. The plaintiff’s expert called the approach circular and outcome-directed.
Perhaps most broadly significant, the trial illuminated the degree to which AI research organizations have operated with governance structures that legal scholars say are poorly matched to the scale of assets and social influence they now command. Professor Amara Lindqvist of the Stanford Law and Technology Institute, who testified as an expert witness, noted that these entities were designed for a world where they managed tens of millions of dollars and a handful of researchers. They now manage tens of billions of dollars and employ thousands of people with significant influence over global infrastructure, she said, adding that the legal frameworks have simply not kept pace.
The reputational stakes extended beyond the two principals. Testimony touched on investment relationships, social dynamics, and hiring practices among a closely networked group of technology executives and researchers, revealing the informal power structures through which large sums of capital and scarce computing resources are allocated in the AI industry.
The jury is expected to begin deliberations next week. Legal analysts said the outcome would likely have implications well beyond the two principal parties, potentially reshaping how state and federal regulators approach oversight of AI research organizations and the complex conversion transactions that several of them have undertaken in recent years. Whichever side prevails, observers agreed that the eleven days of testimony had permanently altered public understanding of how the sector operates behind closed doors.