Elon Musk testified this week in an Oakland federal court, alleging that OpenAI CEO Sam Altman and president Greg Brockman deceived him into providing $38 million in seed funding for a nonprofit that evolved into an $800 billion commercial entity. The trial, centering on the governance of artificial general intelligence (AGI), revealed that Musk’s xAI distills OpenAI’s models to train its own.
This represents more than a billionaire’s grudge match. It is a forensic autopsy of the “nonprofit” myth in the AI era. At its core, the litigation examines whether the pursuit of AGI—AI that can match or exceed human cognitive capabilities—can ever truly coexist with the fiduciary demands of a for-profit corporate structure.
The Distillation Admission: Engineering a Shortcut
The most technically explosive moment of the week occurred when Musk admitted under cross-examination that xAI partly
distills OpenAI’s models. To the layperson, this sounds like a minor technicality. To an engineer, it is a confession of architectural dependency.
Model distillation is a Teacher-Student framework. A massive, computationally expensive Teacher
model (like GPT-4) generates high-quality synthetic data or “soft targets” (probability distributions) that a smaller, more efficient Student
model (like Grok) mimics. By training on the outputs of the superior model, the Student can achieve a fraction of the Teacher’s performance with significantly lower inference latency and reduced hardware overhead on the NPU (Neural Processing Unit) or GPU clusters.
The problem? Most frontier AI labs view distillation as a form of intellectual property theft. OpenAI has previously accused the Chinese firm DeepSeek of this exact practice. By admitting to distillation, Musk has effectively acknowledged that xAI is not building a ground-up alternative to OpenAI, but is instead leveraging OpenAI’s own R&D to accelerate Grok’s scaling.
“Model distillation essentially allows a competitor to skip the most expensive part of AI development—the initial discovery of how to reason—by simply copying the reasoning patterns of a leader.” Dr. Andrew Ng, AI Researcher and Founder of DeepLearning.AI
Musk defended the move as standard practice
for validating AI. However, the industry standard is increasingly shifting toward strict Terms of Service (ToS) prohibitions. This is why Anthropic blocked OpenAI’s access to Claude in August 2025; the moat in the LLM war is no longer just the data, but the behavioral patterns encoded in the weights of the model.
The $1.75 Trillion Valuation Paradox
The courtroom drama highlighted a staggering divergence in valuation and intent. Musk claims he was a fool
who provided free funding
to a nonprofit, only to watch it transform into a behemoth approaching a $1 trillion valuation. He is now asking the court to unwind the restructuring that enabled OpenAI’s for-profit subsidiary.
Yet, the irony is palpable. While Musk argues that OpenAI’s commercialization is a betrayal of humanity, his own AI venture is projected to go public via SpaceX as early as June, with a target valuation of $1.75 trillion.
The financial stakes create a perverse incentive. If the court forces OpenAI back into a strict nonprofit structure, it would likely collapse its current investment model, including the $10 billion investment from Microsoft. Musk’s strategy appears to be a pincer movement: undermine the competitor’s legal foundation while scaling his own valuation through the SpaceX ecosystem.
The 30-Second Verdict: Governance vs. Greed
- The Claim: Musk says he was “baited and switched” into funding a profit-machine.
- The Counter: OpenAI argues Musk simply wants to kill a competitor he no longer controls.
- The Technical Leak: xAI uses OpenAI’s models via distillation, complicating Musk’s “independent” narrative.
- The Risk: A ruling could freeze the IPO path for the world’s most valuable AI entities.
Safety Theatre and the “Terminator” Narrative
Musk’s testimony leaned heavily on existential risk, warning the jury that the worst-case scenario is a Terminator situation where AI kills us all
. He positioned himself as the only adult in the room, citing a conversation with Google cofounder Larry Page who allegedly dismissed the risk of AI wiping out humanity.
Judge Yvonne Gonzalez Rogers was not buying the performance. When Musk’s lawyer, Steven Molo, argued that OpenAI could not be trusted with AI safety, the judge pointed out that Musk is building a company in the exact same space. I suspect there’s plenty of people who don’t wish to put the future of humanity in Mr. Musk’s hands,
she noted.
The contradiction is stark. While preaching safety in court, xAI spent April suing the state of Colorado over an AI law intended to prevent algorithmic discrimination. This suggests that Musk’s definition of safety
is less about preventing a robot apocalypse and more about preventing regulatory friction that slows down his own deployment cycles.
The Recruitment War: Poaching as Strategy
The trial also exposed the brutal talent war within the Silicon Valley AI bubble. Evidence surfaced that Musk actively poached OpenAI employees to fuel Tesla and Neuralink. A 2017 email to a Tesla VP regarding the hiring of Andrej Karpathy revealed Musk’s awareness of the friction: The OpenAI guys are gonna want to kill me. But it had to be done.
This highlights the “brain drain” phenomenon that defines the current AI race. Because the number of researchers capable of managing LLM parameter scaling at the frontier is so small, the competition isn’t just over compute (H100s) or data, but over the specific human intuition required to stabilize training runs.
Musk’s defense—It’s a free world
—is a classic libertarian pivot. But in the context of a lawsuit about “betrayal” and “deception,” the evidence of aggressive poaching paints Musk not as a duped donor, but as a strategic operator who has always viewed OpenAI as a talent incubator for his broader empire.
The Path to AGI: Who Wins?
As the trial moves into its second week, the focus shifts to technical experts. Stuart Russell, a renowned computer scientist at UC Berkeley, is expected to testify on AI safety. His testimony will likely move the conversation away from Musk’s cinematic warnings and toward the actual mechanics of AI alignment.
The outcome of this case will set a legal precedent for how AI companies are structured. If the court finds that a “nonprofit” mission can be legally pivoted to a for-profit model without the consent of original donors, it opens the floodgates for other AI labs to follow OpenAI’s lead. If not, we may see a massive restructuring of the AI industry’s cap tables.
For now, the admission of distillation is the real story. It proves that even the most vocal critics of OpenAI are secretly relying on its weights to build their own future. The “counterbalance to Google” has become the foundation for everyone else.
