America is arguing over the wrong AI obstacle

Last month, two Chinese firms unveiled AI models that garnered little attention outside the industry: Kimi K3 by Moonshot AI and Qwen3.8-Max by Alibaba Group Holding. The market responded before the performance claims were fully verified, resulting in a $3 trillion decrease in chip stock market value. This reaction, rather than the models themselves, is the key story.

The scarcity of expensive, American AI systems was the underlying assumption, but now that assumption is worth less. Washington's instinct is to tighten controls, but the real danger is not runaway machines but the failure to spread American AI technology globally. Restrictionists argue that America's advantage can be maintained through denial, while accelerationists believe it is being stifled by regulation.

However, both camps are arguing about the wrong obstacle. There is little regulation to blame, as the US lacks a comprehensive AI statute. The issue is more about management than technology. Hospitals, insurers, and car plants are not deploying these systems due to bureaucratic hurdles and lack of motivation. These factors are not technological problems but managerial ones, and no export control can fix them.

Cheap, replicable models matter beyond Washington and Beijing, as they allow institutions to download open-weight Chinese models and run them on their servers without buying access to frontier closed-weight American AI models. For most of the world, the question is not which model to license but whose to build on. Washington cannot respond by making American systems harder to obtain, as this would push users elsewhere.

The policy questions are narrow and mundane: can a public hospital acquire an AI system within a year? Do federal agencies have clean enough data to use one? Is there a national standard for validating clinical or credit models to make adoption a defensible decision rather than a career risk? Targeted controls on sensitive chips have a place, but blanket containment does not, and it buys time that nobody is using.

Scarcity is what drove users to seek alternatives. Lastly, it is nearly impossible to force people to uninstall open-weight model apps. The real policy questions are narrow and practical, not speculative or dramatic.