Vulnerability, Robots hijackable by voice command.

The warning is clear.

And if a single word were enough to transform an intelligent robot into a weapon, even when disconnected from the Internet, Chinese researchers have just exposed a worrying flaw in the new generation of commercial robberies, both humanoid and quadrupedal.

In security tests carried out during the GeekCon in Shanghai, specialists demonstrated that some thefts can be completely hijacked using voice commands or short-range wireless connections, without the need for direct access to the network.

The fault is transmitted.

Modern robberies, increasingly autonomous and connected, are more vulnerable than many imagine and invasion can occur in a matter of minutes and in one of the most impressive tests, researchers from the cybersecurity group Dark Navy exploited flaws in humanoid robots equipped with AI agents responsible for interaction and decision making.

A Unitree business model valued at around 100,000 yuan was compromised solely by exploiting loopholes in the voice control system. By circumventing the protections, the researchers gained full access to the machine, in that case, while it was connected to the internet, but the experiment went further; Once kidnapped, the robot began to act as an attack vector, using short-range wireless communication it transmitted the exploit to another nearby robot that was offline and without connection to any network.

Within a few minutes, the second robot was also taken, to demonstrate the real physical risk, the researchers issued hostile command, the robot advanced against a mannequin on the stage and hit it, the message was direct, it is not just about data, it is about bodies, force and movement in the real world.

This type of attack breaks a long-standing assumption of digital security.

The idea of ​​keeping systems offline is enough and a scenario where robots operate in groups, share sensors that communicate locally, a single breach can spread like a domino effect, unlike traditional cyber attacks that cause financial losses or information leaks, hacked robots pose a direct physical risk.

In homes, factories, hospitals or public spaces, a security failure can mean injuries, sabotage even deaths. We already know that thefts can be hacked and as they stop being technological showcases and enter our daily lives, the concern arises: What happens when machines with strength, autonomy and guidance fall into the wrong hands?