OpenAI Co-Founder Greg Brockman Describes Physical Confrontation with Elon Musk

In August 2017, at a Bay Area property Elon Musk's associates called the "haunted mansion," a gathering of OpenAI co-founders turned from celebratory to menacing. Greg Brockman, the company's president and co-founder, told a federal jury on May 5, 2026, that Musk — after learning he would not receive majority control of the AI startup — "stood up and stormed around the table" toward him [1]. "I was sitting in front of the painting and I actually thought he was going to hit me," Brockman testified. "I truly thought he was going to physically attack me" [2].

Instead, Musk grabbed a painting of a Tesla Model 3 that co-founder Ilya Sutskever had given him as a gift, tore it from the wall, and began storming out. Before leaving, Brockman said, Musk turned and demanded: "When are you going to be departing OpenAI?" He added that he would withhold funding until they decided what to do [2].

The testimony, delivered on Day 6 of the Musk v. Altman trial in Oakland, California, has become one of the most vivid moments in a case that could reshape the structure of the most valuable private AI company on Earth.

The Scene at the 'Haunted Mansion'

Brockman's account placed five people at the August 2017 meeting: himself, Musk, Sutskever, actor Amber Heard (Musk's girlfriend at the time), and Musk adviser Shivon Zilis, who now has several children with Musk [3]. The evening started warmly — Musk had recently gifted Tesla vehicles to his fellow co-founders, and Heard poured whiskey as the group discussed OpenAI's future [3].

The mood shifted when the conversation turned to the structural question that would define OpenAI's next decade: how to raise the capital needed to compete in AI development. The co-founders had concluded that a nonprofit structure could not generate the billions required. But when the topic of equity distribution in a proposed for-profit arm arose, Musk demanded a 51% controlling stake and the CEO title [4]. The other co-founders proposed equal shares. Musk declined [2].

No other attendees have publicly offered independent corroboration of Brockman's description of the physical confrontation. Heard and Zilis have not made public statements about the incident. Sutskever, who left OpenAI in 2024 to co-found a rival AI safety company, has not commented.

The Trial: A $134 Billion Reckoning

The confrontation account emerged during what legal observers have called the most consequential tech trial in years. Musk sued OpenAI, CEO Sam Altman, and Brockman in February 2024, alleging they broke commitments to keep the organization a nonprofit and used his $38 million in donations for unauthorized commercial purposes [5]. In a January 2026 filing, Musk's attorneys demanded up to $134 billion in damages from OpenAI and Microsoft, which is also named as a defendant [6]. Musk has since said any damages awarded should flow to the OpenAI Foundation, not to him personally [7].

The trial began with jury selection on April 27, 2026, in Oakland. Musk spent most of the first week on the stand, testifying that OpenAI was founded specifically as a nonprofit to counterbalance Google's growing AI dominance [8]. Under cross-examination by OpenAI attorney William Savitt — who previously represented Musk and Tesla — the billionaire admitted that his own company, xAI, had "partly" trained its Grok model by "distilling" OpenAI's GPT models, a practice that violates OpenAI's terms of service [7]. Musk downplayed the significance, calling it "standard practice to use other AIs to validate your AI" [7].

OpenAI's legal team has argued that Musk "was never committed to OpenAI being a nonprofit" and is suing to undermine a competitor [7].

Source: OpenAI Funding Rounds (Tracxn, Bloomberg)

Data as of Apr 1, 2026CSV

Financial Stakes and the For-Profit Conversion

The amounts at stake dwarf anything in Musk's original $38 million contribution. OpenAI closed its latest funding round in March 2026 at an $852 billion valuation, having raised $122 billion in committed capital [9]. The company completed its conversion to a Delaware public benefit corporation in late 2025, after California Attorney General Rob Bonta approved a restructuring that gave the newly formed OpenAI Foundation approximately 26% of the company's value — about $130 billion — making it one of the most well-endowed philanthropic organizations in the world [10].

The financial transformation of OpenAI's insiders has been equally dramatic. Under cross-examination, Musk's attorney Steven Molo pressed Brockman about his equity stake in OpenAI's for-profit subsidiary, which is currently valued at roughly $30 billion [11]. Molo repeatedly noted that Brockman never followed through on a stated intention to contribute $100,000 — or any cash — to the nonprofit. "I did not end up donating, that is true," Brockman acknowledged [4].

The core financial dispute traces to 2017. Musk wanted majority equity, initial board control, and the CEO role. Brockman and the other co-founders rejected this, arguing that "it was against the mission for any individual to have absolute control over OpenAI" [12]. Musk left the board in 2018. OpenAI established its for-profit arm shortly after [12].

Brockman also testified that Musk told him he wanted control of OpenAI in part to finance building "a city on Mars," which the SpaceX CEO said would require approximately $80 billion [4].

Brockman's Own Credibility Under Fire

If the confrontation account damaged Musk's image, Brockman's personal journal entries — obtained through discovery and read aloud in court — raised serious questions about his own motivations.

In a 2017 entry, Brockman wrote: "Financially, what will take me to $1B?" [11]. In another, he described the public commitment to OpenAI's nonprofit mission as "a lie" [13]. A November 2017 entry recorded Brockman being "warm to steal the nonprofit from [Musk] to convert to b corp without him" [13]. He also wrote: "[Musk's] story will correctly be that we weren't honest with him in the end about still wanting to do for profit just without him" [13].

These entries were consequential enough that Judge Yvonne Gonzalez Rogers cited them in January 2026 when she denied OpenAI's motion to dismiss the case, writing that the entries "suggest Brockman intended to deceive" about the organization's charitable purpose [13].

OpenAI's lawyers have argued the journal entries are "staged for maximum misrepresentation," cherry-picked from hundreds of pages of reflective personal writing [13]. Brockman described the entries as "deeply personal and never intended for public disclosure," calling the experience of seeing them in court "very painful" [14].

The Settlement Texts

Two days before the trial began, on approximately April 25, 2026, Musk texted Brockman to gauge interest in a settlement [15]. When Brockman suggested both sides simply drop their claims, Musk responded: "By the end of this week, you and Sam will be the most hated men in America. If you insist, so it will be" [6].

The judge ruled these texts inadmissible at trial [14]. OpenAI's lawyers had argued the exchange demonstrated Musk's underlying motivation — that "the filing argues the message 'tends to prove motive and bias,' particularly that Musk's objective is to damage a rival AI company and its leadership" [15].

The exchange added to a pattern legal analysts have identified: most assessments characterize Musk's legal case as weak on the merits, with the greater damage potentially inflicted through reputational harm during public proceedings [16].

The Question of Strategic Timing

The confrontation account surfaces at a moment when both parties have strong incentives to shape public perception. OpenAI is preparing for a potential IPO in the second half of 2026, with internal targets pointing to a possible 2027 listing at valuations approaching $1 trillion [9]. Musk's xAI is a direct competitor. The litigation itself could delay or complicate OpenAI's public offering.

Brockman's testimony was delivered under oath, in response to questioning — he did not volunteer the account in a press interview or social media post. The account describes events from 2017, nearly nine years ago, and no independent corroboration has emerged. However, the courtroom setting and the penalty of perjury lend the account more weight than an unsupported public claim would carry.

Musk's side faces its own credibility gaps. His admission that xAI distilled OpenAI's models undercuts his framing as a principled objector to OpenAI's commercial turn [7]. His pre-trial threat to make Altman and Brockman "the most hated men in America" suggests motivations beyond recovering charitable assets [6].

Legal Standards: When Does Intimidation Become Actionable?

Brockman's description — fearing Musk "was going to hit me" — raises a legal question about the line between aggressive negotiation and actionable intimidation.

Under most U.S. jurisdictions, civil assault does not require physical contact; a reasonable fear of imminent harm can be sufficient [17]. Criminal assault or menacing statutes typically require a specific threat to commit a crime that would result in death or serious bodily injury [17]. Brockman's account describes fear of an attack that never materialized — Musk grabbed a painting, not a person.

Employment law sets a different bar: harassment becomes unlawful when conduct is "severe or pervasive enough to create a work environment that a reasonable person would consider intimidating, hostile, or abusive," according to the Equal Employment Opportunity Commission [18]. Whether a single incident of storming around a table during a business meeting meets that threshold is debatable, and no legal action related to the incident itself has been filed.

Workplace psychologists distinguish between high-pressure negotiation tactics — raised voices, dramatic gestures, walkouts — and genuine physical intimidation based on whether the subject reasonably fears bodily harm. Brockman's use of the phrase "I truly thought he was going to physically attack me" places his experience at the more serious end of that spectrum, though Musk's actual behavior (grabbing a painting, leaving the room) remained within the bounds of dramatic exits common in high-stakes business disputes.

Precedents: When Tech Founders Fight

Silicon Valley founder disputes have produced their share of acrimony but rarely allegations of physical intimidation. The most prominent analog — the Zuckerberg-Saverin dispute at Facebook — centered on equity dilution and fiduciary duty, not personal confrontation. Saverin alleged his shares were deliberately diluted from 30% to less than 10%; the case settled out of court in 2009 with Saverin retaining co-founder recognition and signing a non-disclosure agreement [19].

At Uber, founder Travis Kalanick's ouster in 2017 involved board-level power struggles and allegations of a toxic workplace culture, but no public claims of physical intimidation between co-founders [20]. The Theranos case involved fraud allegations against Elizabeth Holmes but similarly did not feature physical confrontation between founders and investors [20].

The Musk-Brockman incident, if accepted at face value, would be unusual in the annals of tech founder disputes — not because high-stakes business meetings never become heated, but because a co-founder has rarely testified under oath that he feared physical violence from another co-founder.

What Comes Next

The trial is expected to continue through May 2026, with Sam Altman's testimony still to come. The jury must determine whether OpenAI's restructuring violated commitments to Musk and its founding mission — a question with implications for the $852 billion company's planned IPO and the broader precedent for how nonprofit-to-for-profit conversions are governed in the tech sector [5].

Both sides have presented damaging evidence against the other. Musk's pre-trial texts and xAI distillation admissions undermine his claim to principled objection. Brockman's journal entries undermine his claim to mission-driven selflessness. The confrontation account adds a human dimension — the image of two men in a room, one fearing violence — to what might otherwise remain an abstract corporate governance dispute over billions of dollars and the future of artificial intelligence.