A debate is growing around how Moonshot’s Kimi K3 reached its current level of performance, after comments from White House science advisor Michael Kratsios suggested the Chinese company copied Anthropic’s Fable model. According to experts cited by TechCrunch, that explanation may be too simple for a model that has improved so quickly.
Researchers speaking to the outlet said Kimi K3’s strength does not look like the result of straightforward distillation alone. Their view is that while model copying or distillation can help transfer capabilities, it would not fully explain such a strong system appearing so soon after Fable. That has led some observers to argue that other training work or additional methods were likely involved.
The discussion also highlights a broader problem in AI: it is often difficult to prove exactly how a large language model was built without direct evidence from the company behind it. As open-weight models become more competitive, questions about whether developers are borrowing from rivals, training on outputs, or independently advancing their own systems are becoming more common.
Kimi K3’s status as a major open-weight LLM has made the issue especially visible. The disagreement between policymakers and technical experts shows how contested AI development has become, particularly when leading models emerge from competing ecosystems and their origins are hard to verify from the outside.
