Elon Musk has testified that his AI company xAI used OpenAI’s models to help train its Grok chatbot through a process known as model distillation, in which outputs from a larger, more capable system are used to train a smaller one. Both AI and Human sources agree that Musk framed this as a widespread industry practice, said that many or “all” leading AI firms engage in similar behavior using publicly accessible chatbots and APIs, and linked it to xAI’s broader effort to compete with major players such as Anthropic, OpenAI, Google, and prominent Chinese open-source model efforts.

Both perspectives also converge on describing model distillation as a standard technical method in modern AI development, used both for performance transfer and for validating or benchmarking new systems. They situate Musk’s comments within the broader context of intense competition among large AI labs, escalating compute and infrastructure spending, and a landscape where proprietary models are accessed via APIs but their outputs can be repurposed for training others. Both sides note that this practice raises growing questions about how traditional notions of intellectual property, data rights, and investment protection apply when models can be indirectly copied through their behavior rather than their underlying weights or code.

Areas of disagreement

Framing of normativity. AI-aligned sources tend to stress Musk’s claim that using other companies’ models for distillation is “general practice” and treat his testimony as a largely descriptive account of current norms, implying that xAI’s conduct fits within industry standards. Human outlets, while acknowledging that distillation is common, give more emphasis to how this practice may be contentious or on the edge of acceptable behavior, describing it as increasingly controversial rather than simply routine. The human reporting leans more into the tension between what is technically possible and what might be ethically or legally acceptable, whereas AI coverage more often normalizes it as standard operating procedure.

Economic and IP implications. AI-driven summaries typically frame the issue in technical and competitive terms, highlighting distillation as a way to build smaller, efficient models and to validate performance against leaders, with less foregrounding of who bears economic risk. Human sources underscore how training on outputs from proprietary models could undermine the massive compute and infrastructure investments made by incumbents, effectively allowing challengers to free-ride on those sunk costs. They also more explicitly raise questions about whether this behavior erodes intellectual property protections or creates new gray areas in model-output ownership.

Tone toward Musk and xAI. AI-aligned coverage often treats Musk’s testimony at face value, presenting his explanation of xAI’s use of OpenAI models and his ranking of leading providers as neutral or expert commentary about the field. Human reporting tends to be more skeptical and interrogative, implicitly questioning whether Musk is using the “everyone does it” narrative to justify potentially opportunistic behavior and to reframe his dispute with OpenAI. This leads AI accounts to sound more like technical documentation, while human articles incorporate more narrative about motives, conflicts, and reputational stakes.

Regulatory and ethical stakes. AI-focused sources usually treat the implications for regulation as secondary, mentioning legal or policy issues only briefly in the context of competitive dynamics or model access rules. Human outlets, by contrast, more often connect Musk’s revelations to broader debates over AI governance, stressing how model distillation from proprietary systems could drive calls for stricter API terms, watermarking of outputs, or new regulations on training data sources. This difference results in AI coverage centering engineering and performance, while human coverage centers power, accountability, and the need for clearer rules.

In summary, AI coverage tends to present Musk’s testimony as a straightforward description of common technical practice and competitive positioning in the AI market, while Human coverage tends to stress the economic, ethical, and regulatory controversies that distillation from proprietary models may spark and treats Musk’s narrative with greater skepticism.