China demo shows one whispered command could let hackers seize robots | The compromised robot used short-range wireless signals to infect another robot that was offline and not connected to any network.
https://interestingengineering.com/ai-robotics/security-flaw-could-allow-hackers-control-robots
5 Comments
“In October, researchers revealed a Bluetooth flaw in Unitree robots that allows wireless root access, allowing a compromised machine to spread attacks to nearby robots and form a dangerous botnet.
In a controlled test, the team demonstrated that a commercially available humanoid robot could be hijacked with nothing more than spoken commands, exposing how voice-based interaction can serve as an attack vector rather than a safeguard.
The robot used in the experiment, a locally manufactured Unitree model costing roughly 100,000 yuan (about US$14,200), was running an embedded large-scale AI agent designed to manage interaction and autonomy. By exploiting a flaw in this system, the researchers bypassed safeguards and gained complete control of the machine while it was connected to the internet.
Once compromised, the robot became an isolated threat. Instead, it was repurposed as a conduit for further attacks. Using short-range wireless communication, the hijacked machine transmitted the exploit to another robot that was not connected to the network. Within minutes, this second robot was also taken over, demonstrating how a single breach could cascade through a group of machines.”
More reasons AI will fail horribly. People are able to prompt their way to free airfare and free car offers from dealerships. They can prompt their way to getting an AI agent to share all customer details. The insurance industry will cause AI to collapse on its own because a bored 12 year old can figure out how to break even the most complicated AI. As someone who has actually implemented an AI agent through salesforce.com; I can assure you that the way it’s set up has more holes than you can believe. Now imagine that 12 year old prompting their way to getting a robot to leave the amazon warehouse and walk 1,000 miles over to his house.
This is the true danger of robot armies, at least with humans your entire army can’t be reprogrammed in an instant to turn around and attack you.
The true danger of robots is that they’re being made by people who still think the “Internet of Things” is still a good idea.
A botnet run by toasters is going to lead an army of pizza-making robots.
This sounds like something a high school kid would do during lab instead of working on his project.