A recent federal court testimony by Elon Musk has stirred the waters in the intense world of artificial intelligence, suggesting that his AI venture, xAI, may have utilized OpenAI’s models for its own training. This revelation emerged during a cross-examination by an OpenAI attorney, part of Musk’s ongoing legal skirmish with the ChatGPT maker. The exchange on the witness stand shed light on a practice known as “distillation,” which is becoming a contentious point among leading AI developers.
During the intense courtroom proceedings, OpenAI’s lawyer, William Savitt, directly questioned Musk about the technique. Savitt inquired if Musk understood distillation, to which Musk replied, “It means to use one AI model to train another AI model.” He then added a critical point, stating that “Generally all the AI companies [do that],” implying this practice is widespread across the industry.
Musk’s Courtroom Revelation on AI Training
Distillation is a sophisticated technique where a smaller, more efficient AI model learns to mimic the performance and behavior of a larger, more powerful one. This process allows the distilled model to run faster and more cheaply, all while retaining much of the superior model’s capabilities. Essentially, it’s a way to transfer knowledge from a complex “teacher” model to a simpler “student” model.
Following this explanation, Savitt pressed further, asking directly if OpenAI’s technology had been employed in any capacity to develop xAI. Musk’s response was succinct, framing the act as a standard industry procedure: “It is standard practice to use other AIs to validate your AI.” These statements, though brief, indicate a potentially significant aspect of xAI’s development strategy and underscore the fierce competition in the AI sector.
Neither OpenAI nor xAI have yet provided official comments on this specific testimony to media outlets. However, this courtroom moment offers a rare glimpse into the competitive tactics and shared technical practices that define the current landscape of AI innovation.
The Controversial Practice of AI Distillation
OpenAI has been actively working to protect its cutting-edge AI models from distillation by competitors, particularly from foreign entities. The company’s concerns were detailed in a February 2026 memo to a House committee, where it stated that it has “taken steps to protect and harden our models against distillation.” This proactive stance highlights the strategic value and intellectual property inherent in these advanced models.
In that same memo, OpenAI explicitly articulated its focus on maintaining a level playing field, stating it was essential to ensure that “China can’t advance autocratic AI by appropriating and repackaging American innovation.” This sentiment underscores a broader geopolitical concern about technological leadership and national security. The potential for rivals to leverage American-made AI without independent innovation is clearly a major worry for the company and the government alike.
Indeed, the Trump administration previously took steps to prevent Chinese companies from distilling American AI models. In an April 2026 memo, Michael Kratsios, then the White House’s director of the office of science and technology policy, indicated that the US government would share information with domestic AI companies regarding foreign distillation attempts. Kratsios publicly reinforced this commitment on X, stating that the “US government is committed to the free and fair development of AI technologies across a competitive ecosystem,” suggesting a balance between open innovation and strategic protection.
Industry Practices and Growing Tensions
Historically, American AI labs have often used each other’s models for various purposes, including benchmarking progress and assessing safety. This collaborative approach fostered a degree of shared learning and collective advancement within the industry. However, the rapidly escalating competitive landscape is significantly altering these norms, pushing companies to adopt more protective stances.
In recent times, some leading AI companies have opted to completely cut off rival labs from accessing their proprietary models. For instance, in August 2025, Anthropic took action against OpenAI, blocking its access to the Claude coding models after alleging violations of its terms of service. This incident set a precedent for more aggressive protection of intellectual property and competitive advantage.
More recently, Anthropic extended this protective measure by cutting off xAI from utilizing its AI models for coding purposes as well. These moves signal a noticeable shift from a potentially more open, research-focused environment to a fiercely competitive arena where access to proprietary AI models is tightly controlled and fiercely defended.
Behind the Legal Battle: Musk vs. OpenAI
The testimony from Elon Musk about xAI’s training practices is just one facet of a broader legal battle and a long-standing rivalry. OpenAI’s lawyer, William Savitt, has been meticulously cross-examining Musk over several days, focusing on his past attempts to assume control of OpenAI and his subsequent mission to outperform the ChatGPT creator.
During the proceedings, Savitt presented a trove of emails and texts dating back to 2017 to bolster his line of questioning. These communications explored whether Musk had exerted pressure on OpenAI by withholding crucial funding and by actively recruiting away key researchers from the organization. The ongoing legal dispute is not merely about a technical detail like distillation; it delves deep into the origins of OpenAI, Musk’s involvement, and the intense, high-stakes competition shaping the future of artificial intelligence.
Source: Wired – AI
