Musk Admits xAI Used OpenAI Models for Training Distillation

Elon Musk testified in federal court that his artificial intelligence startup xAI partially used OpenAI’s models to train its Grok chatbot through distillation techniques. Under oath, Musk acknowledged the practice was “partly” true but characterized it as standard procedure among AI laboratories developing competing systems. The testimony occurred during ongoing litigation involving OpenAI, where training methodologies have become central evidentiary points. Distillation involves generating synthetic data from a larger proprietary model to improve smaller or newer systems, a technique increasingly scrutinized for intellectual property implications. Musk defended the approach by asserting widespread industry adoption rather than denying the specific allegation. Legal experts suggest this admission could influence pending copyright disputes regarding synthetic training data across the sector. Neither party has commented on potential licensing agreements or damages related to the disclosed training methodology. The court record now formally documents xAI’s reliance on competitor outputs during Grok’s development phase.