NVIDIA Launches Halos Full-Stack Safety System to Scale Physical AI Deployment

NVIDIA has launched the Halos full-stack safety system, designed to provide safety assurance across hardware, software, AI behavior, operating environments and the deployment lifecycle for scaled physical AI deployment. ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles by 2035, while Omdia estimates roughly 60 million industrial robots will be deployed between 2026 and 2035. NVIDIA states that Halos is the first and only full-stack safety system for physical AI, helping developers engineer safety across design, validation and deployment. For autonomous vehicles, Halos spans the DRIVE AGX Thor compute platform, Hyperion reference architecture, Halos OS built on ASIL-D certified DriveOS, Alpamayo reasoning vision language action models and a safety evaluation framework. For robotics, it includes the IGX Thor module, Halos Core, Holoscan Sensor Bridge, Isaac Lab and Omniverse simulation, plus the open-source Outside-In Safety Blueprint. Geely, Isuzu, Nissan and Einride are building L4 vehicles on Hyperion; Agility is integrating IGX Thor and Halos Core into the safety system for its Digit 5 humanoid. TÜV SÜD and TÜV Rheinland have conducted independent assessments, and ANAB has accredited the Halos AI Systems Inspection Lab as an ISO/IEC 17020 inspection body. This information comes from NVIDIA's official blog and was organized by AI.