AI & Models
Musk admits xAI used distillation on OpenAI models
Elon Musk testified that xAI partly used distillation techniques on OpenAI models, confirming that American AI labs frequently leverage each other's technology to train their own systems.
On Thursday, Elon Musk testified in a California federal court that his artificial intelligence startup, xAI, used distillation techniques on OpenAI models to train its Grok chatbot. Distillation is the process of training new AI models by prompting publicly accessible chatbots and APIs. When asked on the stand during the trial if xAI had used these distillation techniques on OpenAI’s models, Musk responded, “Partly.” He further asserted that using distillation is a general practice among AI companies, confirming industry assumptions that U.S. developers leverage each other’s systems to avoid falling behind.
The testimony occurred during Musk’s ongoing lawsuit against OpenAI, its CEO Sam Altman, and co-founder Greg Brockman. The legal battle, which went to trial this week in California, features Musk alleging they breached the original nonprofit mission for OpenAI by shifting the entity to a for-profit structure. The trial began this week, featuring testimony from the tech leader on the stand.
The admission of using distillation comes amid broader industry efforts to restrict the practice. OpenAI, Anthropic, and Google have reportedly launched an initiative through the Frontier Model Forum—an industry group focused on AI safety and information sharing—to share information about how to combat distillation attempts from China. These attempts typically involve systematic querying of models to understand their inner workings. While these U.S. companies are working to prevent external users from making suspicious mass queries, Musk’s testimony indicates that similar distillation techniques are actively used internally among domestic competitors.
Later in his testimony, Musk was asked about his previous claims regarding xAI’s competitive positioning. In response, he ranked the leading AI providers, placing Anthropic in the top spot, followed by OpenAI, Google, and Chinese open-source models. He characterized xAI, which started in 2023, as a much smaller operation that currently runs with just a few hundred employees.
Why it matters
Musk’s admission confirms that American AI labs use distillation techniques on each other’s models. This practice threatens the competitive advantage of frontier labs by allowing competitors to create capable models at a lower cost.