NVIDIA announced the launch of NVIDIA Halos for Robotics on June 22, 2026, in Chicago, introducing the industry’s first comprehensive, full-stack safety architecture for physical AI. The framework unifies artificial intelligence compute capabilities with functional safety protocols, addressing a critical bottleneck in scaling autonomous robots for dynamic industrial environments. Built upon NVIDIA’s 18,600-plus engineering years of autonomous vehicle safety development, the system provides developers with a standardized foundation to design, validate, and deploy robotics without compromising operational security. The architecture integrates AI compute, system software, sensor data streams, safety applications, and inspection tools into a cohesive layer. As physical AI transitions from controlled testbeds to active factories, warehouses, and logistics hubs, a unified safety model is essential for compliant human-robot collaboration. NVIDIA leadership emphasized that the framework accelerates development cycles while embedding confidence in system reliability before physical deployment. Agility Robotics has been selected as the inaugural enterprise to implement the technology. Agility will integrate NVIDIA Halos Core software alongside IGX Thor industrial-grade compute modules into its Digit humanoid robot. This integration targets manufacturing, warehousing, and logistics workflows, ensuring that human detection and operational safeguards are hardwired into the robot’s core functionality. The collaboration will utilize the NVIDIA Halos AI Systems Inspection Lab to verify compliance with stringent international standards, including IEC 61508, ISO 13849, and ISO/IEC TR 5469, prior to independent third-party certification. The initiative is supported by a broad industrial ecosystem. Software partners including Acontis, FreeRTOS, and QNX provide real-time operating environments and safety communications. Hardware integrators such as Advantech and NexCobot supply pre-validated IGX-based robotic platforms, while semiconductor manufacturers Infineon, NXP, STMicroelectronics, and Texas Instruments contribute safety microcontrollers and sensor technologies. Industrial application developers like FORT Robotics, Inventec, and KION Group are building functional safety agents using the open-source NVIDIA Halos Outside-In Safety Blueprint. Validation pathways are streamlined through the Halos AI Systems Inspection Lab, which holds ANSI accreditation and maintains recognition from major certification bodies including TÜV Rheinland, TÜV SÜD, UL Solutions, and SGS. This network ensures that compliance testing remains consistent and internationally recognized. NVIDIA Halos Core for IGX is currently available in early access for registered developers operating on Linux and QNX OS for Safety 8.0. The open-source safety blueprint is simultaneously available for public review on GitHub. The release establishes a standardized pathway for accelerating the deployment of certified physical AI systems across global manufacturing and logistics networks.