When Robot Safety Fails Because Its Sensors Are Tricked

Modern robots rely on strict rules to work alongside humans. A factory arm slows down when an operator steps close, a delivery bot halts in a crowded corridor, and a humanoid robot yields when someone crosses its path. In all these cases, the safety programming works as designed, stopping or slowing the machine whenever its sensors report an obstacle.

Researchers at cybersecurity company VicOne are pointing out a major vulnerability: what happens when the safety code runs without error, but the input data feeding it is wrong? If a distance sensor falsely reports an empty hallway, or a stop signal never registers, the machine follows its programming and keeps moving forward, unaware that a person is actually in its path.