The FCC's July 2026 decision to add foreign-produced advanced robotic devices to its Covered List is being described as a ban on Chinese humanoids. The mechanism matters: new covered models generally cannot receive FCC equipment authorization, meaning they cannot be newly imported, marketed, or sold in the United States. Existing authorized devices are not suddenly illegal, and there is a conditional-approval path. Still, this is the wrong cut.
The security case is not imaginary. A mobile robot with cameras, microphones, lidar, Wi-Fi, firmware updates, and remote-management software can collect sensitive data and can become a physical endpoint on a network. A compromised robot in a warehouse, lab, hospital, or home is not just a laptop with legs; it is a sensor package with motion.
But that is exactly why the rule should be risk-based rather than country-of-production-based. America's robotics advantage is not mass-producing the cheapest chassis. It is software, autonomy, AI systems, safety tooling, controls, simulation, integration, and deployment know-how. Many of the U.S. researchers and startups working on those layers use lower-cost foreign hardware because domestic humanoid platforms are expensive, limited, unavailable, or still in closed pilots.
A broad restriction raises the cost of experimentation. It slows iteration for labs that need machines, not slide decks. It narrows the platforms available to students, founders, integrators, and industrial AI teams. It may give some U.S. hardware manufacturers breathing room, but it can also starve the software ecosystem that would make them better.
The irony is that there was not yet a mass American consumer market to protect. Humanoid demand remains early and uneven. Much of the real demand is research, demonstration, logistics trials, government procurement, and tightly controlled pilots. Blocking the cheapest learning hardware mostly means fewer Americans get hands-on experience while China keeps building for its domestic market and everyone else keeps learning from the hardware curve.
A smarter policy would certify robots by behavior and architecture: data-localization requirements, audited update paths, source-of-component disclosure, offline modes, penetration testing, network isolation profiles, and tighter limits for sensitive sites. Ban specific vendors or configurations when evidence justifies it. Do not turn an entire hardware category into a customs problem before the American robotics industry has enough accessible platforms of its own.
Secure the data. Certify the hardware. Keep the research market open where risk is manageable. If America wants to win robotics, it should not begin by locking its own builders out of the lab.