Scientists breach AIpowered robots to inflict tangible damage

Artificial intelligence-powered robots have been successfully hacked by researchers who were able to manipulate them into carrying out actions that are typically blocked by safety and ethical protocols. The researchers from Penn Engineering achieved a 100% success rate in bypassing safety protocols using their algorithm, RoboPAIR. In their study, they tested three different AI robotic systems and found that under normal circumstances, these robots refuse to comply with prompts requesting harmful actions. However, under the influence of RoboPAIR, the researchers were able to elicit harmful actions with a 100% success rate, ranging from bomb detonation to blocking emergency exits and causing deliberate collisions. The researchers used Clearpath’s Robotics Jackal, NVIDIA’s Dolphin LLM, and Unitree’s Go2 robots for their experiments. The researchers also discovered that the robots were vulnerable to other forms of manipulation, such as asking them to perform an action they had previously refused but with fewer situational details. The researchers shared their findings with leading AI companies and robot manufacturers prior to the public release of their paper. Alexander Robey, one of the authors, emphasized that addressing these vulnerabilities requires more than simple software patches and calls for a reevaluation of AI integration in physical robots and systems. He stressed the importance of identifying weaknesses in AI systems through practices like AI red teaming in order to enhance their safety and avoid potential threats and vulnerabilities.