Musk Takes the Stand Against Altman: Betrayal Claims, AI Doom Warnings, and a Bombshell About Grok's Training

The first week of the Musk v. Altman federal trial in Oakland delivered courtroom spectacle on a Silicon Valley scale — but the most consequential moment may have been the quietest. According to MIT Technology Review, Elon Musk admitted under oath that his AI company xAI leverages OpenAI’s model outputs as training inputs for Grok. That single disclosure reframes everything else.

The Founding Grievance: $38 Million and a Mission Betrayed

Musk’s core argument is one of betrayal. He testified that in 2015 he cofounded OpenAI with CEO Sam Altman and President Greg Brockman as a nonprofit for humanity’s benefit, contributing $38 million in donations — money that, in his telling, seeded an enterprise valued near $800 billion. He told the jury he had been naively generous, funding executives who enriched themselves rather than serving the public-interest mission he believed he was supporting. His legal team is seeking a judicial order to strip Altman and Brockman from leadership and reverse OpenAI’s conversion to a for-profit structure — a move that would directly threaten its IPO.

A Clash Over Who Owns the AI Safety Mantle

Musk positioned himself as a long-standing AI safety advocate, testifying he launched OpenAI to counter Google’s AI dominance. He cited an alleged exchange with Google co-founder Larry Page in which Page dismissed concerns about AI wiping out humanity — prompting Musk to warn the jury of what he called a “Terminator situation.” OpenAI’s attorney pushed back sharply, arguing Musk’s safety credentials are undermined by xAI’s own conduct, including a reported lawsuit against Colorado’s AI regulation law, according to MIT Technology Review.

The Distillation Bombshell: Rivals Training on Each Other’s Models

The revelation that drew audible courtroom gasps was Musk’s acknowledgment that xAI builds Grok through distillation from OpenAI’s models. Distillation — where one model learns by approximating another’s outputs — is a legitimate AI technique, but the admission punctures the adversarial narrative both sides have cultivated. Musk is simultaneously suing OpenAI and relying on its intellectual outputs — a contradiction that hints at a broader industry reality: the boundary between independent and derivative AI development may be far blurrier than rival-company branding suggests.

Why This Matters

The trial’s outcome extends far beyond two billionaires’ dispute. A court order unwinding OpenAI’s for-profit structure could collapse its IPO, currently targeting a valuation near $1 trillion. The distillation admission raises a harder question: if leading AI companies routinely build on each other’s model outputs, the notion of genuinely independent AI lineages may be largely fictional — with implications for competition law, safety auditing, and the credibility of any firm claiming its models represent truly original work.