Musk v. OpenAI: The Testimony, the Threats, and the $1 Billion Question

The trial of Elon Musk versus OpenAI opened in a San Francisco federal court on 23 April 2026, and by the second week it had produced the kind of testimony that rarely surfaces outside Silicon Valley boardroom dramas. Greg Brockman, OpenAI's president and one of the co-founders alongside Sam Altman, took the stand and described a confrontation with Musk that left him fearing physical violence. "I thought he was going to hit me," Brockman testified, according to BBC News reporting from 6 May 2026. The statement, delivered under oath in open court, crystallises the depth of personal acrimony between two figures who once shared a common enterprise.

The lawsuit itself is, on its face, a corporate governance dispute. Musk alleges that OpenAI, the nonprofit he helped seed with a reported $44 million contribution in 2015, abandoned its founding mission when it restructured as a commercially capped entity and entered into exclusive cloud-computing arrangements with Microsoft. His legal team argues this shift transformed OpenAI into a de facto subsidiary of a single technology oligarch—Microsoft—which undermined the structure of openness and public benefit that early backers believed they were financing. OpenAI denies the claims, maintaining that its restructuring was necessary to attract the capital required to compete in the accelerating race toward artificial general intelligence, and that the nonprofit parent retains oversight of the commercial subsidiary's activities.

Shivon Zilis and the Board's Complicated Loyalties

Among the more revealing testimonies to emerge from the proceedings is that of Shivon Zilis, a Neuralink executive and the mother of four of Musk's children. Zilis served on OpenAI's board from 2020 to 2023, according to Business reporting published 6 May 2026, placing her at the organisation during the period most directly contested in Musk's complaint. Her relationship with Musk makes her a uniquely complicated witness. She testified in her capacity as a former director, offering insight into how the board processed the restructuring decisions and whether the nonprofit's mission obligations were being honoured.

The Zilis testimony matters for a specific structural reason: it tests the claim that OpenAI's board operated with genuine independence from Microsoft. If a director with deep personal ties to one of the plaintiffs could simultaneously serve on that board and attest to proper governance, it complicates the narrative of captured oversight. The sources do not specify the content of her testimony beyond her formal role, but her presence on the stand reinforces that this case will turn on documents and deliberations that were previously confidential. Board minutes, email chains, and internal AI safety reviews—many of which OpenAI has resisted disclosing—are the evidentiary backbone of Musk's case.

The Threat and the Stakes of Silence

Before the trial began, Brockman testified, Musk delivered a message to both him and Altman that read less like a legal negotiation and more like a warning shot. According to Telegram posts from the ProductHunt and AngelList feeds, both reporting on 6 May 2026, Musk told Altman and Brockman that they would become "the most hated people in America" if they pursued legal claims against him. The phrasing suggests an explicit attempt at reputational leverage—a familiar tactic in high-stakes Silicon Valley disputes, but one that carries particular weight when the defendant is the world's wealthiest individual and one of the most polarising figures in global public life.

The threat, if accurate and properly documented, raises a question that goes beyond the immediate legal dispute: to what extent does the Musk persona function as a coercive instrument in corporate negotiations? The sources do not establish whether this statement has been entered into the court record or whether the defence has challenged its authenticity. But its public emergence fits a pattern in which Musk's communications—tweets, press releases, litigation filings—routinely shape the information environment before formal proceedings can respond. Whether such statements constitute intimidation in the legal sense is a question for the court. Whether they constitute intimidation in the court of public opinion is, given Musk's 200 million-plus following on social media, largely rhetorical.

What We Verified / What We Could Not

The following represents the current state of verifiable evidence as reported across the sourced outlets, alongside the material gaps that remain unresolvable from public sources.

Verified:

• Shivon Zilis served on OpenAI's board from 2020 to 2023, a fact confirmed in Business reporting from 6 May 2026.
• Greg Brockman testified during the second week of the trial, with his statement to BBC News on 6 May 2026 reporting his recollection that "I thought he was going to hit me" during a confrontation with Musk.
• Musk made a statement to Altman and Brockman about becoming "the most hated people in America" if they pursued legal claims, reported by two Telegram sources on 6 May 2026.
• Musk was a founding figure and early funder of OpenAI, with TechCrunch reporting on 6 May 2026 that he contributed approximately $44 million between 2015 and 2018 before departing the board in 2018.
• OpenAI restructured from a pure nonprofit into a hybrid structure with a for-profit subsidiary capped atreturns, a transition that drew Musk's primary legal challenge.

Could not be independently verified from source materials:

• The precise wording or date of the alleged threat made by Musk to Altman and Brockman; the Telegram sources summarise the content but do not include the full text or timestamp.
• The content of Shivon Zilis's testimony beyond her formal role and duration of board service; the Business article does not detail her specific statements.
• Internal board deliberations on the Microsoft partnership and restructuring decisions; these documents have not been released publicly and OpenAI has contested their disclosure.
• Whether Musk's $44 million contribution constituted a binding legal commitment to the nonprofit's ongoing mission, a point central to the case but not addressed in the sourced reporting.

The Structural Stakes: Who Controls the Most Important Technology in the World

Underlying the personal acrimony is a governance question that extends well beyond the parties to the case. OpenAI was founded on the premise that the development of artificial general intelligence—AI systems that match or exceed human cognitive capacity across most domains—should not be controlled by any single entity, whether a corporation or a state. That premise was always in tension with the capital requirements of modern AI development. Training frontier models costs hundreds of millions of dollars per run; the compute infrastructure required to sustain competitive research demands commercial partnerships that inevitably create dependencies.

Microsoft's $13 billion investment in OpenAI, sealed in 2023, is the specific arrangement Musk's lawsuit targets. The deal gave OpenAI access to Azure cloud computing at subsidised rates while granting Microsoft exclusive rights to deploy OpenAI's models through its commercial products. Musk's legal team argues this constitutes a de facto acquisition of a nominally nonprofit entity by a for-profit technology giant, a structure that violates OpenAI's founding charter. OpenAI counters that the arrangement preserves nonprofit control through its parent entity and that the commercial subsidiary's activities remain subject to board oversight.

What is not in dispute is that the outcome of this trial will set precedent for how the AI sector handles the tension between mission-driven nonprofit structures and the capital intensity of frontier research. If Musk prevails, the implications for existing AI labs operating under hybrid structures—Anthropic, DeepMind, Mistral—would be substantial. If OpenAI prevails, it effectively legitimises the current model in which philanthropic origin stories coexist with commercial arrangements that determine who actually controls the technology.

The trial is expected to run through late May 2026. The court has not indicated when it will issue a ruling on the cross-motions currently pending. Until then, the evidentiary record—testimony, documents, and the occasional public statement—will continue to surface in fragments, each adding texture to a dispute that is ultimately about who gets to decide what kind of AI world gets built, and in whose interest.

This desk covered the Musk v. OpenAI trial with primary focus on internal governance and testimony evidence, rather than the broader AI safety or competitive dynamics that dominated wire framing. The personal dimension—Brockman's testimony, the Zilis appearance, the reported threat—received prominence here because those details, while not legally dispositive, illuminate the human architecture of institutional failure in ways that the corporate filings do not.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

• https://t.me/ProductHunt/98432
• https://t.me/AngelList/18923