{"version":1,"type":"story","url":"https://digestai.news/story/vicone-lab-r7-finds-manipulated-inputs-can-bypass-robot-safety-systems","json":"https://digestai.news/story/vicone-lab-r7-finds-manipulated-inputs-can-bypass-robot-safety-systems.json","markdown":"https://digestai.news/story/vicone-lab-r7-finds-manipulated-inputs-can-bypass-robot-safety-systems.md","slug":"vicone-lab-r7-finds-manipulated-inputs-can-bypass-robot-safety-systems","headline":"VicOne LAB R7 finds manipulated inputs can bypass robot safety systems","summary":"VicOne LAB R7 tested how false or crafted inputs—like text on a poster or inaudible audio—can alter robot behavior despite safety protocols. In one test, a robot-dog treated a poster’s text as an instruction, changing its movement. Another simulation using Nemotron on NVIDIA Jetson AGX Orin showed crafted audio altering a hospital-service robot’s behavior. Even independent protective systems may fail if they rely on the same manipulated data, the lab found.\n\nThe research highlights gaps in current safety assessments, which often assume faults are accidental rather than deliberate. For example, a ROS 2/DDS message injected into a stationary robot triggered movement, and acoustic attacks could distort gyroscope readings in humanoid balance controllers. VicOne warns that cyber threats require reassessing risk models, testing sensor redundancy, and validating safeguards against adversarial inputs. The lab’s tools—Radeis for pre-deployment validation and Rthena for operational monitoring—help teams address these risks across a robot’s lifecycle.","keyPoints":["VicOne LAB R7 demonstrated how crafted text or audio can alter robot behavior despite safety rules","Manipulated inputs bypassed protective systems when they shared the same faulty data source","New standards like China’s GB/T 45502-2025 and EU Machinery Regulation require testing against deliberate attacks"],"whyItMatters":"Robots rely on sensor data to trigger safety actions, but adversarial inputs can fool these systems. This research forces industry to update safety testing to include deliberate manipulation, not just accidental faults.","category":{"slug":"robotics","name":"Robotics & Physical AI","url":"https://digestai.news/category/robotics"},"entities":{"companies":["VicOne","NVIDIA"],"models":["Gemma 4 E4B","Nemotron"],"people":[]},"firstPublishedAt":"2026-10-01T15:02:19Z","updatedAt":"2026-10-01T15:02:19Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"The Robot Report","title":"Your Robot’s Safety Functions Already Work. What If the Input Lies?","url":"https://therobotreport.com/your-robots-safety-functions-already-work-what-if-the-input-lies","publishedAt":"2026-10-01T15:02:19Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"VicOne LAB R7 finds manipulated inputs can bypass robot safety systems\", 1 October 2026, https://digestai.news/story/vicone-lab-r7-finds-manipulated-inputs-can-bypass-robot-safety-systems","publisher":"Digest AI","title":"VicOne LAB R7 finds manipulated inputs can bypass robot safety systems","datePublished":"2026-10-01T15:02:19Z","url":"https://digestai.news/story/vicone-lab-r7-finds-manipulated-inputs-can-bypass-robot-safety-systems"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}