On March 25, 2026, on the red carpet in the White House East Room, Melania Trump appeared alongside Figure 03. Facing the cameras, the robot said: “I am a humanoid robot made for America.”

Less than 24 hours later, on March 26, bipartisan leaders in the U.S. Senate jointly introduced the American Safe Robotics Act, proposing to blacklist Chinese robots from government procurement.

The timing was almost too precise. From that moment, the EU, South Korea and the United States had effectively formed a broad front to block Chinese robots.

Can this campaign to choke off Chinese robotics ultimately succeed?

Blocking Rationale

The American Safe Robotics Act immediately set off a storm.

The official rationale is that the measure is meant to protect U.S. national security, data privacy and reduce technology risks.

The bill alleges that Chinese robots are equipped with multimodal sensors, including vision and audio systems.

It says they could send data back to China in sensitive political, military and infrastructure settings. The bill even alleges that Chinese robots all contain “backdoors” and could be remotely hijacked by China.

The bill also accuses China of low-price dumping. “With state support, Chinese companies are expanding rapidly, seizing global market share at low prices, and threatening the U.S. robotics industry, AI R&D, supply chain security and technological leadership.”

On that basis, it proposes a sweeping ban on Chinese robots.

But the bill also leaves an opening: the U.S. military, intelligence agencies and the Department of Defense can apply separately to procure Chinese robots.

The condition is that they must be physically disconnected from the internet and tested in closed environments. In plain terms: ordinary government agencies are banned from touching them. Military think tanks can buy them freely, take them apart, study the technology and copy whatever they want.

“In public, they kick you out of the market and stop you from making money; in private, they strip your technology down and use it themselves.”

Luo Shen, a veteran robotics researcher, put it bluntly. Chinese internet users were furious: “Classic hypocrisy: on one hand they say you’re unsafe, on the other they want to buy you and study you inside out.”

American internet users also criticized the move, calling it “the most shameless protectionism in history.” Luo Shen said the bill had been in the works in the U.S. for more than a year.

Step by step, with targeted strikes.

Targeted Strikes

The story dates back to 2024.

At the time, Unitree robotic dogs appeared in footage from a China-Cambodia joint military exercise. Although Unitree clarified that it does not sell products directly to the military, the footage still drew the attention of the U.S. Congress.

In May 2025, 24 members of the U.S. House of Representatives signed a letter listing Unitree’s supply records to universities with defense backgrounds and accusing it of “participating in military-civil fusion projects and receiving state funding.”

During that period, research institutions piled on. First, they claimed Unitree’s Go1 had a remote access channel that allowed external parties to connect directly to and control the robot.

Later, security firm Alias Robotics published a paper claiming that Unitree’s G1 robot automatically sent data back to Chinese servers every five minutes. Whether these allegations against Unitree hold up is not the point.

U.S. sanctions followed almost immediately.

In February 2026, the U.S. Department of Defense placed Unitree directly on the “1260H list,” an official roster of companies deemed linked to China’s military-industrial complex. Just one month later, the American Safe Robotics Act was introduced.

Industry insiders said: “Whether the bill ultimately passes is not the main issue. Once the U.S. Congress formally raises security risks, government procurement departments in Europe, Japan and South Korea are highly likely to follow.”

That trend is already spreading in sync. On the same day the bill was introduced, South Korea announced anti-dumping duties of up to 19.85% on Chinese industrial robots, forcing the landed price of Chinese equipment up by nearly 20%.

The EU moved even earlier. In June 2025, it used the International Procurement Instrument, or IPI, for the first time to shut Chinese medical device companies out of government procurement contracts worth more than 5 million euros.

That procurement restriction framework for Chinese high-tech products could be extended to robotics at any time.

The blockade against Chinese robots by the United States, South Korea and the EU is now taking shape.

The Rise of Chinese Robots

Luo Shen said the U.S. move is also an indirect admission of the strength of Chinese robotics.

At a U.S. House hearing on March 17, 2026, Max Fenkell, head of policy at Scale AI, acknowledged a reality: U.S. AI models are indeed powerful, but when it comes to building bodies for AI and collecting data in the physical world, China has already pulled ahead.

Global shipments of humanoid robots reached about 14,500 units in 2025, with Chinese companies taking nearly 90% of the market.

Unitree and AgiBot alone sold more than 10,000 units combined. In the U.S., the star lineup of Tesla Optimus, Figure AI and Agility together shipped only about 450 units.

Beyond the 20-fold gap in sales, the price gap is another source of anxiety for Americans.

In humanoid robots, Unitree’s G1 is priced at $21,600 and has won market recognition. Figure 03 has a target price of roughly $20,000, but it has not entered mass production, and its actual cost remains unclear.

In quadruped robots, Boston Dynamics’ Spot sells for $74,500. Unitree’s Go2 costs just $1,600. The budget for one American robot dog can buy 46 Chinese ones.

Fenkell has also analyzed the reason behind this overwhelming cost advantage: the data collection cost for Chinese AI robots is more than 60% lower than in the United States.

This was not created by subsidies alone. It is the result of more than a decade of supply chain building in China, backed by national-level attention and mobilization. Take Unitree: its self-developed share for core components such as motors and sensors exceeds 90%.

Because it controls its own components, the hardware can be priced extremely low. Because the hardware is cheap enough, it can be deployed at massive scale. Because robots are everywhere, real-world AI data can be collected at very low cost.

That has created a fast-spinning positive flywheel. Unable to beat China in mass production or cost, the U.S. has returned to an old playbook: label the threat, then change the rules.

Scale AI has proposed classifying Chinese robotics data as a “national security threat,” and the Trump administration has signed an executive order promoting the U.S. AI technology stack. Can they really block the rise of Chinese robots?

The Higher the Wall, the More It Blocks the Builder

History has already offered an answer. At the end of 2024, the U.S. House passed the National Defense Authorization Act with the “Countering CCP Drones Act” included, targeting Chinese drones directly. DJI still holds more than 70% of the global market.

If drones could not be stopped, robots are likely to follow the same path.

Chinese companies are no longer sitting at the same card table waiting to be hit. Just six days before the U.S. Safe Robotics Act was introduced, on March 20, 2026, Unitree Technology, a leading domestic robotics company, had its STAR Market IPO application accepted. It plans to raise 4.202 billion yuan, with annual revenue above 1.7 billion yuan and a gross margin as high as 60%.

Sanctions cannot sever a supply chain where Unitree’s self-developed share of core components exceeds 90%, nor can they shut off a vast demand market.

If hardware cannot be contained, what about software?

That is exactly where the competition is fiercest right now. A robot’s brain is different from AI software. It cannot get smarter only by reading text online. It has to go into real environments, pick up cups, climb stairs, and build experience frame by frame through countless real collisions and trial-and-error cycles.

On this front, China is using its vast deployment scenarios to “feed” AI.

Beijing’s Shijingshan district has built a humanoid robot training ground of more than 10,000 square meters. Hubei has set up an innovation center with 23 simulated scenarios. Multiple institutions in Shanghai have jointly released a dataset containing data from one million real machines.

Even Max Fenkell, policy head at U.S. data-labeling giant Scale AI, admitted at a congressional hearing that Chinese robots can collect data for as many as 17 hours a day in large warehouses, while the U.S. currently lacks real-world deployment scenarios at comparable scale.

The more the U.S. keeps Chinese robots out, the more it hurts itself.

Robots become smarter the more they are used, and their data comes from running in real environments.

Industry insiders believe that by keeping cost-effective Chinese hardware out, the U.S. may appear to be protecting domestic manufacturing, but in practice it is also cutting itself off from the world’s richest source of real-world scenario data: China.

The result of America’s wall-building is that it blocks itself. The higher the wall, the wider the efficiency gap will become.

From drones to robots, congressional bans have never been a cure-all.

Chinese companies still hold an overwhelming advantage in shipments. That the U.S. Congress is rushing to pull the plug now is, in itself, the strongest validation of China’s robotics supply chain on the eve of its breakout.

This time, however, China’s well-resourced robotics industry may not pay the price for someone else’s anxiety.