At a federal court hearing held in California this Thursday, Elon Musk stated that his AI company, xAI, had leveraged models from OpenAI to improve its own system performance. This statement centers on "model distillation," an issue of heightened sensitivity within the current AI industry. Model distillation refers to using a large-scale model as a "teacher" to transfer its capabilities into smaller models, thereby achieving a balance between performance and efficiency. While widely adopted internally by companies—for instance, training lightweight versions of proprietary large language models to reduce costs—the technique has also drawn accusations against some firms for rapidly replicating competitors' abilities through distilling rival models. During cross-examination regarding whether xAI performed distillation on OpenAI's technologies, Musk did not directly confirm but remarked that "almost every AI company does it." Under further questioning, he added that "some do so." He emphasized that validating one's own models using other AI systems constitutes standard practice across the sector. In recent years, intensifying competition among large models has blurred the boundaries surrounding distillation techniques, sparking widespread controversy. Multiple institutions, including Anthropic and OpenAI, have accused certain entities of exploiting distillation methods to acquire their model capabilities. Google has already implemented measures to counteract alleged "distillation attacks," viewing them as potential intellectual property infringements. Nevertheless, the industry generally regards distillation itself as a neutral technology. In a related statement, Anthropic noted that while the method can be used to build more efficient commercial models, it may also enable rivals to replicate advanced capabilities at low cost; its legitimacy depends entirely on specific usage patterns and compliance boundaries.