On July 24, NVIDIA CEO Jensen Huang made his debut on X, sharing a joint open letter titled Open Weights and American AI Leadership, co-signed by over twenty major technology firms, startups, and investors. The letter advocates for open-weight AI models, arguing they enhance cybersecurity, accelerate innovation, and preserve technological sovereignty. Huang’s first public post ignited the Kimi Panic, a widespread market reaction triggered by Moonshot AI’s July 16 release of Kimi K3, a 2.8-trillion-parameter open model featuring breakthrough coding capabilities that approach the performance of leading proprietary systems. The release has intensified a geopolitical and commercial contest between open and closed AI ecosystems. OpenAI and Anthropic have formally alleged that Chinese developers are using model distillation to illegally extract intellectual property, prompting US Treasury and Commerce departments to explore sanctions and targeted restrictions on Chinese open-weight models. The Little Tech Association, a coalition of nearly 200 startups and investors, warns that banning affordable open models would force American companies into expensive API dependency, potentially bankrupting capital-constrained firms. At the core of the dispute is model distillation, a long-established machine learning technique where a smaller student model learns from the outputs of a larger teacher model. Industry experts clarify that distillation transfers behavioral patterns and reasoning capabilities rather than copying weights or source code. While OpenAI and Anthropic openly utilize distillation internally and for customer fine-tuning, they now seek to restrict third-party extraction via strict API terms of service. Legal scholars note that copyright enforcement remains ambiguous, with current constraints governed more by commercial contracts than clear statutory frameworks. Technically, distillation complements rather than replaces foundational training. Chinese teams have demonstrated significant algorithmic and engineering autonomy, developing proprietary architectures, mixture-of-experts routing, and efficient training pipelines. However, restrictions on closed-API access could slow downstream fine-tuning, self-distillation, and automated evaluation processes, particularly for smaller developers lacking in-house teacher models. Meanwhile, US policymakers are reassessing whether to expand existing semiconductor export curbs to include direct restrictions on model weights and inference services, though enforcing such measures against globally distributed open weights remains technically unfeasible. The controversy underscores a broader industry shift: as open models close the performance gap with proprietary systems, the battle over training data, compute access, and licensing terms is redefining AI competition. With Kimi K3’s weights scheduled for full release on July 27, global developers are already drafting deployment protocols. The distillation debate will likely persist in legal and policy forums, but the diffusion of open-weight architectures continues to accelerate, forcing both Silicon Valley and Washington to adapt to a decentralized AI ecosystem where capabilities, once unlocked, cannot be easily contained.