Musk Admits xAI Used OpenAI Models for Training Under Oath
Original: Elon Musk Seemingly Admits xAI Has Used OpenAI's Models to Train Its Own
Why This Matters
Reveals competitive dynamics in AI industry where companies use rivals' models for training
During federal court testimony, Elon Musk confirmed xAI has used OpenAI's models through distillation to train its own AI systems. Under cross-examination, Musk called it standard practice among AI companies while defending against OpenAI's lawsuit.
In federal court testimony Thursday, Elon Musk acknowledged that his AI company xAI has used OpenAI's models to train its own systems through distillation. When asked by OpenAI lawyer William Savitt whether xAI had distilled OpenAI models, Musk responded 'partly' and claimed 'generally all the AI companies' engage in this practice. Distillation involves using a larger AI model to train a smaller, faster version. Musk also confirmed OpenAI technology was used 'to validate' xAI's AI, calling it standard practice. The admission came during ongoing litigation between Musk and OpenAI. OpenAI has actively worked to prevent competitors, particularly Chinese lab DeepSeek, from distilling its models, stating in February 2026 it had 'taken steps to protect and harden our models against distillation.'