Greg Brockman, the president of OpenAI, recently shared a dramatic story about a heated confrontation with Elon Musk. The incident took place during a meeting at Musk’s estate in 2017, and it highlights the intense disagreements behind the scenes of one of the world’s leading AI companies. Brockman’s testimony offers a rare glimpse into the tensions that shaped OpenAI’s early days and Musk’s role in its development.

The 2017 Mansion Meeting and Tense Exchanges

In August 2017, Brockman and fellow OpenAI cofounder Ilya Sutskever visited Musk’s sprawling estate south of San Francisco. The meeting was supposed to be a discussion about the future of OpenAI. During the gathering, Musk’s then-girlfriend, Amber Heard, served whiskey to the group before leaving with a friend. Brockman recalls Musk giving them new Tesla Model 3 cars as a gesture, which he interpreted as an attempt to win their favor.

However, the atmosphere quickly became tense. The core issue was how to structure OpenAI’s funding and control. Musk wanted a for-profit arm to attract investment, but Brockman and Sutskever opposed giving Musk full control, fearing it would turn the company into a “dictatorship.” After some back-and-forth, Musk rejected their proposal for shared control. Brockman described the moment as frightening, recalling that Musk stormed around the table and Brockman feared he might physically attack him. Musk then took a painting from the room, threatened to cut off funding, and left, but not before his chief of staff, Shivon Zilis, called to suggest the story wasn’t over.

Behind the Scenes and Musk’s Behavior

Later that night, Brockman and Sutskever learned that discussions about their future with Musk continued. Zilis, who was an adviser and friend, kept them informed of Musk’s updates for years afterward. Brockman described Zilis as a proxy for Musk at times, especially when she joined OpenAI’s board in 2020 and later had Musk’s twins in 2021. Brockman only learned about the children through news reports, despite her telling him about the pregnancy beforehand.

During his testimony, Brockman emphasized that Musk’s unpredictable behavior and desire for control were problematic. He recounted how Musk dismissed early AI prototypes, calling them “stupid,” which demoralized the researchers who had worked hard to develop technologies that would eventually power products like ChatGPT. Brockman believed Musk lacked the vision and patience needed to lead an AI company focused on innovation and collaboration.

He also shared that Brockman, Sutskever, and CEO Sam Altman considered removing Musk from the OpenAI board during ongoing negotiations about creating a for-profit subsidiary. Although they discussed various options, Brockman and Sutskever ultimately decided it was wrong to oust Musk. He left the company voluntarily in early 2018 after expressing doubts about OpenAI’s future success, according to an email he sent at the time.

Throughout the years, Zilis remained involved, acting as a bridge between Musk and OpenAI. Brockman noted that her role was complex, balancing her loyalty to Musk with her responsibilities at OpenAI. Despite internal disagreements, she helped manage Musk’s frustrations, which Brockman said was crucial for keeping the company afloat. However, her departure from the board marked a turning point, and Brockman’s account underscores the challenges of managing a high-stakes AI venture led by a famously unpredictable founder.

Inspired by

• https://www.wired.com/story/greg-brockman-testifies-elon-musk-fight-trial/