Musk's lawsuit against OpenAI begins around its first week.

      Three days of cross-examination in Oakland revealed some awkward admissions related to a $130 billion lawsuit, including that xAI utilizes OpenAI’s models for training. Ultimately, it will be the judge, rather than the jury, who makes the decision.

      Elon Musk took the stand in Oakland on Tuesday, sharing a narrative he has recounted for the past two years. He claimed to have founded OpenAI in 2015 to prevent advanced artificial intelligence from falling into the hands of a single entity. According to him, his former collaborators Sam Altman and Greg Brockman secretly transformed the lab into a for-profit organization, secured billions from Microsoft, and excluded him from the process. Musk framed the 2024 lawsuit he filed as an attempt to restore the original nonprofit structure and reclaim what he believes was taken from him.

      However, after three days, that narrative seemed significantly more disputed than it did at the beginning of the week.

      Musk's case faced "some rough spots," including his own admissions during cross-examination, the judge's repeated cautions regarding the scope of the dispute, and pre-trial rulings that constrained the claims being presented. Collectively, these factors have made the billionaire's case against the world's leading AI company appear, at least at this point, more challenging to win than initially suggested.

      The trial commenced on April 28 in the federal courthouse in Oakland, overseen by Judge Yvonne Gonzalez Rogers. A nine-member jury was selected the day prior, with Musk, his legal team, OpenAI, Altman, Brockman, and Microsoft all involved in the lawsuit.

      The primary damages claim exceeds $130 billion, although some early reports have estimated it at $150 billion; regardless, the structural changes Musk is pursuing, such as partially reversing OpenAI’s transition to a for-profit model, are arguably the more significant demands.

      The procedural setup is atypical. Although a jury has been convened, its verdict will be merely advisory. Judge Gonzalez Rogers will make the ultimate determination regarding liability and remedies, with an expected ruling by mid-May.

      As such, the trial serves more as a lengthy public deposition in front of the judge rather than a classic contest for the jury's support, especially since she has already refined the case to some extent. She preemptively dismissed Musk's fraud allegations and cautioned both parties.

      Over the course of three days, Musk was the primary witness, undergoing questioning first from his own attorneys, followed by rigorous cross-examination from William Savitt, OpenAI’s lead attorney.

      The first uncomfortable moment arose concerning the commitment to nonprofit principles that is central to the case. Savitt presented internal documents suggesting that Musk had advocated for OpenAI to pivot towards a for-profit model under his leadership in 2017 and 2018, and claimed he distanced himself from the initiative when that did not happen. Musk disputed this portrayal but acknowledged the existence of the documents.

      The second moment of discomfort, audible from the audience, was Musk’s admission that his AI company, xAI, which developed the Grok chatbot, relies on OpenAI’s models for training, effectively using outputs from the very system he argues was wrongfully privatized.

      The third issue was procedural, with Savitt contending that Musk had delayed too long in filing the lawsuit and that some key claims were submitted after the statute of limitations had expired. Whether the judge accepts this defense remains to be seen, but the timeline will be entered into the record.

      Even before the trial began, Judge Gonzalez Rogers had already altered the landscape of the case. Her pre-trial rulings dismissed Musk's fraud claims and narrowed the focus to whether OpenAI violated its charitable trust and contractual obligations during its restructuring.

      This adjustment simplifies the case and makes it more straightforward to litigate, although it likely complicates Musk's ability to win based on his original narrative of betrayal.

      On the third day, Gonzalez Rogers advised the lawyers to avoid treating the trial as a referendum on AI safety or Altman's character, a tendency both sides have been prone to. Musk reiterated his longstanding concerns about the existential risks posed by AI, a response the judge seemed to view as irrelevant to the legal questions concerning OpenAI’s directors and their fiduciary duties.

      What lies ahead?

      The trial is anticipated to last another two to three weeks. Altman, Brockman, Microsoft CEO Satya Nadella, and several of OpenAI's earliest engineers are all expected to testify. Musk's expert witnesses, as outlined in court documents, include Berkeley AI researcher Stuart Russell and Columbia Law School's David Schizer, a tax and nonprofit expert.

      OpenAI plans to present its own lineup of governance and AI safety experts, with reports indicating that the defendants aim to highlight Grok's safety record to the jury.

      Musk could potentially regain some ground. Cross-examinations of his co-founders might yield significant admissions, and the extensive documentary evidence, which both parties acknowledge spans thousands of