For years, American labs held an uncontested lead in artificial intelligence. That is no longer obviously true: Chinese open-weight models have closed much of the gap, and by some measures have pulled ahead. The stakes are not merely commercial. Whoever leads sets the terms for what the world runs on, who gets access to it, and what is built into the models themselves, including (in the Chinese case) political censorship and reported evidence that these systems may hand certain users deliberately weakened code. Yet American firms are adopting them anyway, sometimes for cost, sometimes because our own models are constrained from doing the work. So what would losing this race actually cost us? Why are American companies turning to Chinese models to defend their own infrastructure? And can Washington regulate a technology that outpaces the law faste ...