Musk Admits xAI Used OpenAI Models to Train Grok in Federal Testimony

Elon Musk confirmed in federal court this week that his AI startup xAI relied on OpenAI's models to train Grok, his company's competing chatbot—a disclosure that adds a layer of irony to his ongoing legal battle against the very firm he helped found.

Musk made the admission while testifying in a California courtroom, where he was questioned about model distillation. The technique involves a larger AI model serving as a "teacher" to pass knowledge to a smaller "student" model. While the practice is common within companies developing sequential AI versions, critics argue it can also allow smaller labs to replicate the performance of more advanced competitors by mimicking their outputs. Musk confirmed that xAI employed OpenAI's models in this capacity to improve Grok's capabilities.

The testimony comes as Musk pursues a lawsuit against OpenAI, alleging the company has strayed from its original nonprofit mission. The revelation that xAI leveraged OpenAI's technology to advance its own product signals potential competitive and intellectual property tensions that could factor into the broader legal dispute. It also raises questions about the practical boundaries of model distillation within the AI industry—whether the practice crosses lines when a direct competitor uses another's models to train its own systems. The case continues to unfold against the backdrop of intensifying competition in the AI sector, where access to training data and model architecture advantages have become central to commercial battles.