Humanoid robots have long occupied an awkward space in the public imagination: more viral video curiosity than serious industrial force. They stumble, they struggle with basic manual tasks, and they remain largely absent from real workplaces and homes. That makes the Federal Trade Commission’s recent decision to ban foreign-made advanced robots, including humanoids, quadrupeds, and wheeled robots, a genuinely surprising move. Not because the technology warranted it, but because of what the decision signals about how the United States government now thinks about AI.
A Trade Playbook Extended Into New Territory
The FTC’s ruling rests on two stated justifications. The first is national security: robots operating in homes and sensitive facilities collect substantial amounts of data, and foreign-made machines could pose surveillance risks. An FTC document released alongside the ruling cited a real incident in which a single individual was able to gain remote control of 7,000 robot vacuum cleaners, illustrating how networked physical devices can become security vulnerabilities at scale.
The second justification is economic: protecting a domestic robotics supply chain from Chinese competition. This logic is not new. The United States has applied similar reasoning to solar panels, electric vehicles, and drones, using tariffs and procurement rules to slow the adoption of cheaper Chinese alternatives. The trade-offs are always the same: higher prices for consumers and researchers in exchange for a more insulated domestic industry.
What is new here is the sector being protected. Robotics, in this framing, is no longer a standalone hardware industry. It is now treated as a frontier of AI, and therefore subject to the same protective instincts the administration has applied to leading AI laboratories. The FTC’s decision should be read alongside reports that the administration is also considering restrictions on open-source Chinese AI models, which often rival those from companies like OpenAI and Anthropic at a fraction of the cost. Blocking access to those models would, according to figures cited in the source, eliminate an estimated $25 billion in annual savings for businesses. The robotics ban follows the same logic: limit access to cheaper foreign alternatives to give domestic players room to grow.
The Research Problem Nobody Planned For
Here is what most coverage of this ban misses: the US robotics industry is not yet capable of replacing what it is being protected from. American robotics companies and university research labs depend heavily on affordable Chinese hardware to conduct the experiments that drive the field forward. Robots learn new tasks by doing them repeatedly, across large fleets, and that requires buying many machines at low cost.
The price gap is not marginal. A four-legged robot from Unitree, China’s leading humanoid robotics company, costs around $4,600. A comparable robot from Boston Dynamics can cost close to $278,000. That is not a competitive disadvantage. It is a structural barrier to research at scale.
Aaron Prather, director of market intelligence for the Association for Advancing Automation, a robotics trade group, has noted that the ban “creates a challenge for US humanoid researchers,” describing Chinese models as offering the best price-to-capability ratio available. An internal review conducted by his organization found that 90% of recent robotics research papers from US universities relied on robots made by Unitree. A policy designed to strengthen the domestic robotics sector could, in practice, slow the research pipeline that sector depends on.
The asymmetry between the two industries is stark. Unitree is preparing to go public with a valuation targeting nearly $6 billion. No US robotics company offers a meaningful comparison. Figure’s humanoids are not yet selling at scale. 1X’s robots are not yet shipping to homes. Google recently announced a new AI model designed to help humanoids learn tasks faster, with tying a trash bag cited as a notable demonstration of progress. That detail is not a joke. Given how difficult fine motor control remains for robotic hands, it represents genuine advancement. But it also illustrates how early-stage this technology still is.
What It Means to Treat Robots as Strategic AI Infrastructure
The deeper significance of this ban is not about robots at all. It is about a shift in how governments are beginning to define the boundaries of AI policy.
For years, AI policy debates centered on software: large language models, training data, compute access, algorithmic accountability. Physical robots seemed like a separate conversation. The FTC’s ruling collapses that distinction. If a humanoid robot is essentially an AI system with a body, then controlling who makes the body is an extension of controlling the AI stack itself.
This framing has real consequences for researchers, companies, and anyone thinking about how AI capabilities develop. Restricting access to affordable hardware does not just affect prices. It shapes which experiments get run, which research questions get answered, and ultimately which teams accumulate the practical knowledge needed to build the next generation of systems. Gavin Kenneally, CEO of Ghost Robotics, a company that makes four-legged robots for inspection tasks, has argued that the cybersecurity risks from foreign-made robots are genuine and that a more level competitive environment benefits both customers and the industry. That perspective is reasonable. But it sits in tension with the reality that the domestic industry is not yet positioned to absorb the demand it is being handed.
In Short
The FTC’s ban on foreign-made robots is not primarily a robotics story. It is an AI policy story. The administration is extending its protective posture beyond software and leading labs to cover the physical infrastructure of AI, including a sector that is still finding its footing. The risk is that a policy designed to accelerate domestic robotics ends up slowing the research that would make domestic robotics competitive in the first place.
Based on reporting from MIT Technology Review.