Sen. Tom Cotton and Senate Minority Leader Chuck Schumer introduced the American Security Robotics Act last Thursday. The bill would prohibit federal agencies from buying or operating unmanned ground vehicles made by Chinese companies — a broad category that includes wheeled patrol platforms, quadruped bots like the Unitree Go2, and the humanoid robots that have become the public face of the technology competition between the U.S. and China. The sponsors cited national security risks: data harvested by robots and sent back to China, systems that could be remotely accessed or disabled. The framing was familiar — China is flooding the American market with technology that poses espionage risks. What the senators did not mention: those risks had already been documented, publicly, months ago.
In September 2025, cybersecurity researchers at Alias Robotics published a full technical audit of the Unitree G1 humanoid robot — a machine already being used in laboratories and some police departments. Their findings were not ambiguous. The G1 uses Bluetooth Low Energy to handle initial Wi-Fi setup, a process protected by a single hardcoded encryption key that is identical across every Unitree robot ever manufactured. Break one, break all of them. The researchers, working with colleagues Andreas Makris and Kevin Finisterre who developed a separate exploit they called UniPwn, found that the BLE vulnerability could be exploited wirelessly to achieve root-level access — the highest possible privilege on the device. From there, an attacker could inject arbitrary code that executes with full system authority every time the robot starts up. The September 2025 audit demonstrated the full attack chain; Unitree later said it had completed most fixes, though the current patch state of deployed units was not clear as of the bill's introduction.
What makes UniPwn particularly uncomfortable is what happens after infection. The researchers demonstrated that an infected robot could scan for other Unitree robots in BLE range and automatically compromise them, creating a robot botnet that spreads without user intervention. Makris told IEEE Spectrum the researchers asked themselves what would happen if an attacker implanted themselves into one of these police dogs. The Nottinghamshire Police in the UK had been testing a Unitree Go2 quadruped; Makris tried to notify them before publication. They did not respond.
The data exfiltration finding is harder to explain away as an engineering oversight. The Alias Robotics team found that the G1 transmits telemetry to servers in China every five minutes — audio, visual, and spatial data — without users' knowledge or consent. The company did not disclose this in any public documentation. Mayoral-Vilches told IEEE Spectrum that Unitree ignored repeated outreach attempts from security researchers over the course of months before the disclosure. Unitree's public response, posted to LinkedIn in late September, said the company had "completed the majority of the fixes" and would roll them out "in the near future." As of the bill's introduction last week, it was not clear what had been fixed, or whether units already deployed had received updates.
Cotton and Schumer named no specific companies in their public statements. The Reuters report cited Agibot and Unitree as the Chinese firms preparing IPOs in China this year whose products have attracted U.S. attention. Agibot and Unitree declined to comment for this article.
The legislation would exempt U.S. military and law enforcement agencies from the procurement ban if the robots in question cannot transmit or receive data from China — a carve-out that acknowledges the agencies already have Chinese robots and may need them for study and reverse-engineering. Rep. Elise Stefanik introduced a companion bill in the House.
The Association for Uncrewed Vehicle Systems International endorsed the legislation, calling it "an important step to reduce national security risk by restricting U.S. government procurement of uncrewed ground vehicles of all types from foreign adversaries." AUVSI's language signals the scope extends well beyond the humanoid form factor: the group represents the broad autonomous systems industry, and its support suggests the bill has backing beyond pure geopolitics.
The timing is politically resonant. On the same day Cotton and Schumer announced the bill, Melania Trump walked alongside a Figure AI humanoid robot at an education summit at the White House, a staged moment designed to showcase American robotics capability. Figure has a $39 billion valuation [source needed] and a pilot deployment arrangement with BMW. The contrast — Washington promoting domestic humanoid robots on the same day it moved to restrict Chinese ones — was not lost on observers.
This is the second time in recent months Washington has moved to restrict Chinese autonomous systems from federal use. The drone ban enacted last year took a similar form: federal agencies were prohibited from operating Chinese-made drones, with limited exemptions for existing inventory. The Chinese drone manufacturers did not disappear from the U.S. market; they found workarounds, including labeling changes and third-party intermediaries. The same dynamic will likely play out here. But the robotics version has an additional complication: the vulnerabilities have been publicly documented, which means agencies that continue using these systems are doing so with full knowledge of the risks.
The bill does not yet have a committee assignment in either chamber. Its path to a vote is unclear. What is clear is that the problem it addresses — Chinese robots inside American institutions, with documented pathways for remote access and data exfiltration — is not a future scenario. It is the present.
