Your Robot’s Safety Functions Already Work. What If the Input Lies?

- Your robot’s safety functions already work. What if the input lies? Robots rely on distance readings and stop signals, but what happens when those inputs are manipulated? At VicOne LAB R7, we found that adversarial images can make vision-language models ignore instructions, while crafted audio alters simulated hospital robots. Even a simple text poster changed a robot dog’s movement. Safety systems depend on untrusted data; if a gyroscope is spoofed by sound, balance controllers correct for non-existent tilts. Deliberate attacks turn unlikely faults into repeatable exploits. Redundant sensors fail if one manipulation affects multiple checks. As China’s GB/T 45502-2025 and the EU Machinery Regulation demand security evidence, teams must test against cyber threats, not just accidental errors. Security strengthens safety. Whoa, that's mega-illegal. I'll swap that node in twelve minutes to fix it.