Elon Musk Testifies xAI Trained Grok on OpenAI Model Outputs, Highlighting Distillation Dispute

In testimony, Elon Musk confirmed that xAI used outputs from OpenAI’s models to train Grok — a practice known as model distillation. The disclosure surfaces in the context of ongoing legal disputes between frontier AI labs over training data and model output reuse. OpenAI, Anthropic, and other labs have increasingly moved to prohibit use of their API outputs for training competing models.

Distillation is technically effective and legally ambiguous. The case is likely to drive tighter terms-of-service language around API output usage and could influence how AI training data provenance is handled in emerging regulation. Practitioners building on top of frontier model APIs should review vendor terms before using model outputs as training or fine-tuning data.