USC Professor Zhang Huyue: Legal regulation is becoming the invisible "sixth layer of the cake" that allows China to surpass the US's hardware and software advantages in physics AI.

Mars Finance reported on June 24th that Professor Zhang Huyue of the Gould School of Law at the University of Southern California (USC) pointed out at the 2026 Summer Davos Forum that the implementation of Physical AI is extremely dependent on legal infrastructure, which is fundamentally different from generative AI. China, with its centralized and efficient system, has significantly lowered the barriers to road testing of autonomous driving through "regulatory sandboxes" and pilot policies. Although Baidu started six years later than Waymo, it has surpassed Waymo in mileage thanks to wider road testing permits and has expanded to more than 20 cities. In contrast, the US, due to state/city level legal constraints and the public's extremely low tolerance for machine errors, has only opened up to six cities. She emphasized that China is reducing hardware costs and activating the data flywheel through unified standards (such as humanoid robot interface specifications) and government-led data infrastructure. Meanwhile, strict cross-border data transfer controls (such as the ban on automakers exporting data), while limiting Tesla's FSD training in China, have built a "data moat" for local companies like Xiaomi. Legal regulation is becoming an invisible "sixth layer of the cake" for China's Physical AI to surpass the US's hardware and software advantages.