OpenAI's Greg Brockman Recalls Heated Confrontation with Elon Musk Over Company Control

In a dramatic testimony during the ongoing trial of Musk v. Altman, Greg Brockman, cofounder and president of OpenAI, recounted a tense 2017 meeting with Elon Musk at his Hillsborough estate. The gathering, which also included OpenAI’s cofounder Ilya Sutskever and Musk’s then-girlfriend Amber Heard, was ostensibly to discuss the future of the nascent AI research firm. However, Brockman testified that Musk had attempted to "butter them up" by gifting them new Tesla Model 3 cars, seeking to establish a sense of indebtedness before pushing for a for-profit arm and, crucially, absolute control over the company. Brockman and Sutskever strongly objected to granting Musk what they perceived as a "dictatorship" over AI development, proposing instead a model of shared control. This rejection led to a heated confrontation. Brockman vividly recalled Musk standing up and storming around the table, stating, "I actually thought he was going to hit me, physically attack me." According to Brockman, Musk then grabbed a painting Sutskever had made for him, declared he would cut off funding to the nonprofit until they quit, and stormed out. Yet, later that night, Musk’s chief of staff, Shivon Zilis, called to reassure them that "it's not over," signaling ongoing discussions. These events, according to OpenAI, are emblematic of Musk's erratic behavior that undermines his current legal claims. Musk alleges that his $38 million in donations were misused as OpenAI transitioned into an $852 billion for-profit entity, known for products like ChatGPT. Brockman, however, offered another anecdote to illustrate why he believed Musk was unfit to control an AI company. He described how Musk vehemently dismissed an early AI chatbot prototype, calling it "stupid" and demoralizing researcher Alec Radford to the point of nearly quitting the field. Brockman emphasized that Musk's inability to see the potential in this foundational technology, which later underpinned ChatGPT, highlighted his lack of necessary vision. Discussions about potentially voting Musk off the OpenAI nonprofit board continued for months as negotiations for a for-profit sibling company dragged on. Despite initial thoughts of removing him, Brockman and Sutskever ultimately decided it felt "wrong." Musk eventually departed on his own in early 2018, deeming OpenAI to be on a path of "certain failure." Shivon Zilis, acting as a proxy for Musk, continued to keep him informed about OpenAI's developments, even joining its board in 2020. However, Zilis herself became a source of boardroom conflict. After giving birth to Musk’s twins in 2021, Brockman learned of Musk's paternity through news articles, not directly from Zilis, who claimed it was via IVF and entirely platonic. This led to calls from several board members for her removal. Brockman and Sutskever, however, convinced the board to let her stay, believing she helped manage Musk’s frustrations. Zilis eventually left the board in 2023 after Musk launched his rival AI lab, xAI. Other board conflicts, including Brockman's support for the removal of Quora CEO Adam D’Angelo and AI safety researcher Helen Toner, further underscored the internal struggles at OpenAI.