Tengrinews.kz – A humanoid robot developed by Chinese company Unitree sparked controversy after a demonstration where it appeared to go “out of control” and attack a man, setting off heated discussions about the safety of future technologies.
According to witnesses, the H1 robot began moving erratically during a public test, even knocking over a computer that controlled it. The incident quickly went viral on social media, with users claiming the robot had “gone crazy” and “attacked” an engineer. It was eventually stopped by cutting its power supply.
Footage claimed to show a Unitree H1 (Full-Size Universal Humanoid Robot) going berserk, nearly injuring two workers, after a coding error last week at a testing facility in China. pic.twitter.com/lBcw4tPEpb
— OSINTdefender (@sentdefender) May 4, 2025
However, Unitree clarified that the robot didn’t actually malfunction in the way many feared. The company explained that H1 was attempting to maintain balance in unfamiliar conditions. The robot is equipped with a stabilization system designed to help it adjust to shifts in its posture or terrain.
During the demonstration, the robot was secured with a safety tether attached to its head - a standard precaution for public events. Ironically, the tether was the very thing that triggered the erratic behavior.
“The robot’s stabilization algorithms weren’t designed to handle restrictions from a fixed point at its head. It interpreted the tether as a destabilizing force,” the company explained.
At first, H1 tried to correct its posture with small adjustments, but as it failed, the movements became sharper and more exaggerated in an attempt to “break free” from the constraint.
Footage from the event shows the robot thrashing wildly, which some viewers likened to scenes from sci-fi movies where machines turn on humans. Comments online even referenced SkyNet, the rogue AI from the Terminator franchise.
Experts noted that despite appearances, the incident wasn’t a sign of sinister robot rebellion but rather highlighted the challenge of adapting control algorithms to real-world variables.