Elon Musk Admits xAI Used OpenAI Models To Train Grok

During a federal court testimony in California this week, Elon Musk confirmed that his artificial intelligence startup, xAI, utilized models from rivals OpenAI and Google to refine its own chatbot, Grok. Musk characterized the practice as "distillation," a process where a developer uses a larger, more established model to help train or calibrate an aspiring one. This admission came during a legal battle regarding Musk's departure from OpenAI, the company he co-founded before launching his own competing ventures.

The revelation is significant because Musk has frequently criticized OpenAI for its transition from a non-profit foundation to a profit-driven commercial entity. Using a competitor's proprietary data is a common but controversial practice in the AI industry, often restricted by terms of service that forbid using outputs to build competing software. Musk’s testimony suggests that despite xAI’s rapid growth and independent data sources like the X social media platform, it still relied on established industry leaders to bridge the developmental gap.

Moving forward, legal experts and industry analysts will be watching to see if this admission impacts Musk's broader litigation against OpenAI or sparks further intellectual property disputes. While xAI recently secured billions in funding to build its own massive supercomputer, the reliance on "distillation" highlights the immense difficulty of building a top-tier AI model entirely from scratch. The case continues to pull back the curtain on the aggressive and often overlapping development cycles of the world's most powerful tech companies.

This story was originally reported by The Verge.