In August 2017, a pivotal meeting took place at Elon Musk’s lavish estate in Hillsborough, California, involving two key figures from OpenAI, Greg Brockman and Ilya Sutskever. The mansion, which Musk described as “haunted,” served as the backdrop for discussions about the future direction of OpenAI and its potential for-profit arm. During the gathering, Musk’s then-girlfriend, Amber Heard, played the role of host, serving whiskey before departing with a friend.
As the meeting got underway, Musk showcased his generosity by gifting both Brockman and Sutskever with new Tesla Model 3 cars. Brockman, who serves as OpenAI’s cofounder and president, recounted this moment in federal court during the trial for Musk v. Altman. “It felt like he was buttering us up,” Brockman observed, suggesting that Musk aimed to foster a sense of indebtedness. In a gesture of appreciation, Sutskever presented Musk with a painting of a Tesla.
The meeting focused on the need to establish a for-profit segment to attract billions in investment for computational resources. However, tension arose as Musk insisted on retaining control over the company. Brockman and Sutskever pushed back against this notion, fearing it would lead to a “dictatorship” over artificial intelligence development, advocating instead for a shared governance model.
Despite the collaborative spirit, Musk was unwilling to compromise. According to Brockman’s testimony, Musk suddenly stood up, pacing the room, with Brockman fearing a potential physical confrontation. In a dramatic turn, Musk seized the painting and declared he would cut off funding to OpenAI unless Brockman and Sutskever resigned. He left the meeting abruptly, but not before Shivon Zilis, Musk’s chief of staff, reached out to both cofounders later that evening, indicating that discussions were still on the table and that “it’s not over.”
This intense exchange underscores a series of concerning behaviors attributed to Musk, which OpenAI believes undermine his claims regarding the misuse of his substantial financial contributions to the organization. Musk has asserted that his donations, totaling approximately $38 million, were mishandled, leading to the emergence of the now-thriving $852 billion for-profit entity known for innovations like ChatGPT and Codex. Meanwhile, Brockman, alongside OpenAI CEO Sam Altman, has denied any misconduct, with jury deliberations in the Musk v. Altman case potentially commencing as soon as next week.
After concluding his testimony, Brockman reflected on the challenges of meeting one’s idols. According to OpenAI attorney William Savitt, while Brockman had held Musk in high regard for his business prowess, his relentless desire for control became a source of concern. On the other side, Musk’s attorney, Marc Toberoff, voiced that the crux of the matter lies in Brockman’s motivations for seeking shared control, which were scrutinized during the trial.
In a further attempt to highlight Musk’s unsuitability for leading an AI enterprise, Brockman recalled an incident involving an early AI chatbot demonstration. Musk had expressed harsh criticism of the chatbot’s capabilities, calling it “stupid” and suggesting that a child on the internet could perform better. Radford, the OpenAI researcher behind the chatbot, was left demoralized, and Brockman, along with Sutskever, invested significant effort in rebuilding Radford’s confidence. Brockman argued that Musk’s inability to recognize the potential in nascent technology ultimately disqualified him from holding sway over OpenAI’s future. “You needed to dream a little bit,” he remarked, insinuating that Musk had failed to do so.
During the lengthy negotiations regarding the for-profit initiative, Brockman, Sutskever, and Altman briefly considered ousting Musk from the board of OpenAI’s nonprofit. They frequently met over whiskey to discuss alternate funding strategies. Although they identified what they did not want to pursue, a consensus on actionable strategies proved elusive. Ultimately, they refrained from removing Musk, feeling it was morally dubious. Musk eventually distanced himself from OpenAI, citing his perception of the organization’s trajectory as one of “certain failure,” as expressed in an email from early 2018.
Shivon Zilis, who served as an adviser to both OpenAI and Musk, remained a key link between the two camps in the following years. Brockman referred to her as a “friend” and noted that he had initially met her in 2012 or 2013, underscoring the intricate dynamics at play within the rapidly evolving landscape of AI development.
