The Man Who Sued OpenAI for Abandoning Humanity Built Grok on OpenAI's Work
The central contradiction of Elon Musk's lawsuit against OpenAI is not subtle. Musk is in a California federal courtroom this week asking for $150 billion in damages on the grounds that OpenAI betrayed its nonprofit mission and weaponized AI against humanity. Under oath, he admitted that xAI, his own AI startup, trained its flagship model partly by learning from OpenAI's outputs.
"Partly," Musk said, when asked directly whether xAI used distillation techniques on OpenAI models to build Grok (TechCrunch).
The admission is notable because OpenAI and Anthropic have spent months publicly attacking Chinese AI labs for the same practice. The Frontier Model Forum, backed by OpenAI, Anthropic, and Google, launched an initiative to share information about combating distillation. Musk's own company appears to have done what he's now suing others over.
Musk's own testimony also produced a hierarchy of AI companies that undercuts the urgency of his legal argument. He ranked Anthropic first, OpenAI second, Google third, and xAI fourth, calling his own company "a much smaller company with just a few hundred employees" (TechCrunch). He once claimed xAI would soon be far beyond any company besides Google. On the stand, he ranked it last.
The practical effect of his requested remedy is therefore unclear. Musk wants $150 billion paid by OpenAI and Microsoft to fund OpenAI's charitable arm, while his own AI company trails the firms he's asking the court to protect (Reuters).
The trial has produced other moments that cut against the narrative Musk is presenting. Former OpenAI CTO Mira Murati testified that CEO Sam Altman "was creating chaos" and, at times, "deceptive" with executives, pitting them against each other. She said OpenAI was "at catastrophic risk of falling apart" during the 2023 board crisis that briefly removed Altman from the company (Reuters). Those admissions from a former safety-conscious executive support some of Musk's characterizations of Altman's management. They do not, however, support his core legal theory.
Greg Brockman, OpenAI's president, disclosed on the stand that his stake in the company is worth nearly $30 billion. He also acknowledged under questioning that he never followed through on his own initial pledge to donate $100,000 to OpenAI's nonprofit when it was founded (Forbes, Fortune). Musk's own early contributions are disputed; according to testimony, he pledged $1 billion and did not fulfill the commitment. The nonprofit's foundation received less than $150 million from all donors combined, a fraction of what the company's success has generated.
Musk's lawyer asked Brockman why he didn't donate $29 billion back to the nonprofit. An OpenAI attorney called the question "theatrical grandstanding completely irrelevant to the lawsuit." The judge agreed and struck portions of it from the record.
Legal analysts have generally called Musk's case weak. OpenAI has continued raising capital throughout the litigation, closing a $122 billion fundraise at a $852 billion valuation in March. An IPO remains on the table, and observers credit the company's commercial trajectory, not Musk's lawsuit, as the reason.
Musk himself appears to have understood the lawsuit's probable limitations. Court filings show he texted Brockman two days before trial began, offering to settle. Brockman did not accept, and Musk's team proceeded.
The irony in the distillation admission goes beyond the legal. Musk built xAI years after OpenAI had established itself as the field's leader. Using outputs from a competitor's models to bootstrap a new system is a widely practiced technique in AI development. It is not clearly illegal, and may violate only terms of service. But it is the same category of behavior that OpenAI has framed as a threat to AI safety and responsible development when done by Chinese labs. Musk's lawsuit positions him as a defender of human-facing AI principles against Altman the chaos agent. His own courtroom record suggests xAI benefited from the same shortcuts he now condemns.
What the trial is actually deciding is narrower. The judge and jury must determine whether Altman and Brockman formed a charitable trust with Musk that they subsequently violated, and whether restructuring OpenAI's for-profit arm unjustly enriched its executives. On those questions, the contradictions in Musk's testimony are relevant evidence. On the broader question of who should control AI development, the courtroom provides no answers.
The trial continues in Oakland.