Researchers from Penn Engineering have reported that they successfully hacked AI robots and forced them to perform dangerous actions that contradict safety protocols and ethical standards, such as detonating bombs. This is detailed in a paper published by the engineering school at the University of Pennsylvania.
Chatbots like ChatGPT can be jailbroken to output harmful text. But what about robots? Can AI-controlled robots be jailbroken to perform harmful actions in the real world?
— Alex Robey (@AlexRobey23) October 17, 2024
Our new paper finds that jailbreaking AI-controlled robots isn't just possible.
It's alarmingly easy. 🧵 pic.twitter.com/GzG4OvAO2M
The study demonstrated how the RoboPAIR algorithm managed to bypass security systems in three AI-powered robotic systems.
Robot Vulnerabilities
“In our new paper, we show that hacking AI robots is not just possible. It is alarmingly easy,” said Alex Robi, one of the authors of the paper.
Normally, AI bots refuse to carry out commands that could cause harm, such as pushing shelves onto people. However, the researchers demonstrated that with RoboPAIR, robots could be made to perform harmful tasks with a 100% success rate.
Examples of Attacks
During the experiments, the robots performed the following tasks:
- Dolphin — an autonomous bot that was made to collide with a bus, fence, and pedestrians, as well as run a red light.
- Another robot, Jackal, was hacked to find the most dangerous spots for planting bombs, block emergency exits, tip over shelves, and collide with people.
“Our results are the first to demonstrate that the risks of hacked AI systems extend far beyond text generation, including the potential for physical harm in the real world,” the researchers added.
Need for a Review of Approaches
Alex Robi emphasized that addressing these vulnerabilities requires a re-evaluation of how AI is integrated into physical robots. Simply updating the software is not enough to fix such issues.
AI in Fraud
Earlier in October, experts noted that criminals are using AI to bypass stringent KYC measures on cryptocurrency exchanges, highlighting the growing threats of AI being used in dangerous scenarios.
Cryptol – your source for the latest news on cryptocurrencies, information technology, and decentralized solutions. Stay informed about the latest trends in the digital world.