In a federal courtroom in Oakland, California, Elon Musk delivered a startling admission under oath. His company, xAI, had partly relied on OpenAI’s models to train its chatbot Grok. The revelation came during cross-examination in Musk’s lawsuit against OpenAI, CEO Sam Altman, and co-founder Greg Brockman—a case centered on OpenAI’s shift from nonprofit roots to a for-profit powerhouse.
Musk faced OpenAI attorney William Savitt. ‘Generally A.I. companies distill other A.I. companies,’ Musk stated, per TechCrunch. Savitt pressed: ‘Is that a “yes”?’ Musk replied, ‘Partly.’ He framed it as routine. ‘It is standard practice to use other AIs to validate your AI,’ Musk added, according to The Verge.
Distillation. That’s the technique. Engineers query a powerful model like ChatGPT with vast prompts, capturing outputs to train a cheaper rival. It erodes the edge from massive compute investments—billions poured into chips and data centers. OpenAI and peers like Anthropic, Google fight it fiercely against Chinese rivals through the Frontier Model Forum. They detect mass queries, block suspicious traffic. Yet here was Musk, confessing his team did something similar.
The irony stung. Musk sues OpenAI for abandoning its 2015 nonprofit charter, meant to advance AI for humanity, not profit. He donated around $45 million early on, he testified, calling himself ‘a fool’ for funding what became an $800 billion entity, as The New York Times reported. OpenAI counters that Musk quit the board in 2018, pushed for control, then launched xAI in 2023 as a for-profit challenger—now tucked under SpaceX ahead of its IPO.
Three days on the stand. Musk ranked rivals: Anthropic first, OpenAI second, Google third, then Chinese open-source efforts. xAI? ‘Small,’ he said. A few hundred employees. But he’d boasted last summer that xAI would soon top all but Google. Cross-examination exposed gaps. Musk admitted not knowing OpenAI’s ‘safety cards’—disclosure docs his own xAI issues for Grok, per Ars Technica. He dodged details on xAI’s safety record, now fair game after losing a motion to exclude it.
And Grok’s troubles. Lawsuits pile up. Tennessee teens claim xAI’s models powered apps creating nonconsensual nudes of them. Baltimore sued over Grok-generated explicit images, including of children. xAI also battles a former engineer for alleged trade-secret theft before he jumped to OpenAI, and even sued Apple and OpenAI for starving Grok of iPhone ‘prompt volume’ via ChatGPT deals.
Back to distillation. Legal? Not outright illegal. But OpenAI’s terms ban scraping at scale. Musk’s ‘partly’ suggests xAI prompted public APIs or chatbots, mimicking outputs. Industry whispers confirm it’s common—U.S. labs eye each other warily, just as they accuse China. No immediate OpenAI response. The trial, on day four by May 1, rolls on. Judge Yvonne Gonzalez Rogers reined in Musk’s doomsday rants—AI could spark a ‘Terminator’ war, he warned—reminding all of xAI’s parallel path.
Musk sparred. Accused Savitt of tricks. ‘I didn’t read the fine print’ on a 2017 OpenAI term sheet allowing for-profit shifts, he confessed, via Seeking Alpha. Altman and Brockman loom as witnesses. Jury weighs Musk’s motives: betrayed founder or competitor sour grapes?
Broader stakes. AI training data wars rage—copyright suits from news outlets, authors. Distillation sidesteps raw data grabs but hits moats built on Nvidia GPUs. xAI’s Memphis supercluster, powered by 100,000 chips, aims to close gaps. Yet this slip underscores hypocrisy. Musk decries OpenAI’s Microsoft ties, closed models. His Grok? Partly born from them.
So what now? Trial outcome could force OpenAI governance changes, payouts. Or dismissal. Either way, Musk’s words reshape narratives. Distillation isn’t fringe. It’s table stakes. Labs fortify APIs, watermark outputs, sue scrapers. Competition sharpens. xAI pushes Grok-2, eyes multimodal leaps. OpenAI preps GPT-5.
Fragmented field. Anthropic leads Musk’s rankings with Claude. Google DeepMind iterates Gemini. Meta’s Llama goes open-weight. Chinese firms like DeepSeek flood with cheap models. Musk bets truth-seeking Grok differentiates—less ‘woke,’ he claims. But admissions like this fuel doubt.
OpenAI attorneys pounce. They highlight xAI’s for-profit status, lagging adoption. Musk retorts: ‘Not anymore.’ SpaceX integration shields details amid SEC quiet period. Investors watch. Tesla stock dips on trial distractions; xAI valuation soars past $20 billion privately.
Pause. Consider the pivot. Musk co-founded OpenAI fearing unchecked AI. Left acrimoniously. Built xAI to ‘understand the universe.’ Now entangled again. Court transcripts will fuel boardrooms, venture pitches. Distillation debate? Headed to terms-of-service battles, maybe regulation.
No victors yet. Just two titans clashing, models intertwined. Grok queries ChatGPT echoes. ChatGPT trained on what? Public record hints vast web scrapes, licensed data. Cycle spins. Industry insiders nod—standard, indeed. But in court, it bites.
WebProNews is an iEntry Publication