Elon Musk Confirms xAI Used OpenAI Models for Grok
Elon Musk stated under oath that his company xAI has relied on OpenAI's models to enhance its own systems. The revelation occurred Thursday in a California federal courtroom amid a high-profile dispute involving OpenAI.
Musk, founder of xAI, addressed questions about model distillation. This technique uses a bigger AI model as a guide to teach a smaller one. He confirmed it was partly true that xAI applied this method with OpenAI's technology to boost performance.
Details from the Testimony
During cross-examination, the judge inquired if Musk understood model distillation. Musk replied that it means employing one AI model to train another. When directly asked if xAI had distilled OpenAI's technology, Musk noted that all major AI firms engage in such activities. Pressed further, he clarified, "Partly."
Musk added that checking one's AI against others is routine. "It is standard practice to use other AIs to validate your AI," he said. xAI launched Grok, its chatbot, as a competitor to models from OpenAI and others. Musk started xAI in 2023 to challenge what he views as dominant players in AI development.
The testimony forms part of ongoing live coverage from the trial between Musk and OpenAI CEO Sam Altman. Musk co-founded OpenAI in 2015 as a nonprofit focused on safe AI. He departed in 2018 amid disagreements and later sued, claiming OpenAI shifted to profit-driven goals against its original mission.
Understanding Model Distillation
Model distillation serves as a key tool in AI training. A large model transfers knowledge to a compact version, making it faster and less resource-heavy. Companies often use it internally to create efficient variants for users.
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
However, the practice draws scrutiny when applied across firms. Smaller labs sometimes mimic rivals' capabilities quickly and cheaply. Legal boundaries remain unclear, often hinging on terms of service or intellectual property rules.
OpenAI has voiced worries about firms like DeepSeek copying its work. Anthropic pointed to DeepSeek, Moonshot, and MiniMax in similar claims. Google has implemented defenses against what it terms distillation attacks, describing them as theft of proprietary technology.
Broader Industry Tensions
Anthropic detailed its stance in a blog post. The company noted distillation works well for internal use, such as producing affordable models for clients. Yet it warned that rivals exploit it to gain advanced features without full independent effort.
"Distillation is a widely used and legitimate training method," Anthropic wrote. "For example, frontier AI labs routinely distill their own models to create smaller, cheaper versions for their customers. But distillation can also be used for illicit purposes: competitors can use it to acquire powerful capabilities from other labs in a fraction of the time, and at a fraction of the cost, that it would take to develop them independently."
These concerns have grown as AI competition intensifies. Musk's xAI positions itself against OpenAI, which he accuses of straying from open-source roots. The trial probes OpenAI's structure and practices, with Musk's admission adding a new layer to the proceedings.
xAI released Grok in late 2023, integrating it with Musk's X platform. OpenAI, backed by Microsoft, leads with ChatGPT. The feud highlights rivalries in a field racing toward advanced general intelligence.