Nvidia Unveils Halos for Robotics — The Tech Aimed at Making Humanoid Robots Safe to Work Beside

As humanoid robots move out of research labs and onto factory floors, warehouse aisles, and hospital corridors, one question has loomed over the entire industry: how do you actually prove one of these machines is safe to work beside? Nvidia believes it has built the answer. The chipmaker has announced Halos for Robotics, which it describes as the industry's first full-stack safety system purpose-built for physical AI — a combination of specialized hardware, software, and an accredited testing and certification programme designed to give humanoid robot makers, and the companies that deploy them, a credible path to proving their machines won't injure the humans around them. The announcement represents one of Nvidia's most concrete steps yet toward addressing what has consistently been identified as the single biggest barrier to widespread humanoid robot adoption: not capability, but trust. What Halos for Robotics Actually Includes Halos for Robotics is built across three integrated layers, according to Nvidia's announcement. At the hardware level sits NVIDIA IGX Thor, an industrial-grade AI computing platform built with functional safety in mind, paired with Holoscan Sensor Bridge for high-speed sensor connectivity — the combination that allows a robot to process camera, lidar, and other sensor data fast enough to detect and react to a human entering its workspace in real time. Above that sits Halos OS, a dedicated software stack that handles the actual safety functions and applications a humanoid needs to operate responsibly — things like emergency stop logic, safe-speed limiting when humans are nearby, and the kind of redundant fail-safe behaviors that industrial safety standards typically require. The third and arguably most novel piece is the NVIDIA Halos AI Systems Inspection Lab, which Nvidia describes as the world's first ANSI National Accreditation Board (ANAB)-accredited program specifically for functional and AI safety testing in physical AI systems. In practice, this gives Nvidia's robotics partners a structured way to prepare their safety systems for formal third-party certification by major certification bodies including TÜV Rheinland, UL Solutions, TÜV SÜD, exida, SGS, and CertX — the organizations that ultimately issue the safety certifications industrial customers require before they'll deploy a robot on their floor. "As AI-enabled robotics moves into industrial environments, the industry needs standardized, internationally recognized frameworks to assess safety across increasingly complex systems," said Laurie E. Locascio, president and CEO of ANSI, commenting on the accreditation. "ANAB's accreditation of the NVIDIA Halos AI Systems Inspection Lab confirms the program has the competence and impartiality to evaluate robotic AI systems against recognized safety requirements, giving companies a rigorous and internationally recognized foundation for their path to certification." Agility Robotics Is First in Line Humanoid robotics company Agility Robotics has become the first partner to integrate elements of Halos for Robotics into its own proprietary safety system, which it uses to power its Digit humanoid — a robot already deployed in real-world industrial settings for customers including Amazon, GXO, Schaeffler, and Toyota Motor Manufacturing Canada. Specifically, Agility is integrating NVIDIA IGX Thor and Halos Core into its existing safe human-detection system. IGX Thor provides the industrial-grade AI compute with built-in safety capabilities, while Halos Core handles the software layer responsible for the robot's safety-related operating functions. Agility and Nvidia plan to use the Halos AI Systems Inspection Lab specifically to verify that Digit's safety-related software, AI components, and cybersecurity protections meet rigorous, internationally recognized industrial standards — including IEC 61508, ISO 13849, and ISO/IEC TR 5469 — before pursuing final third-party certification. "For humanoids to deliver value at scale, safety has to be built into the robot and validated across the entire system," said Peggy Johnson, CEO of Agility, underscoring a point that has become something of an industry mantra as humanoid robots move closer to mainstream commercial deployment. Why This Matters Now The timing of Halos for Robotics is no accident. Nvidia has spent the first half of 2026 aggressively building out what amounts to a full operating environment for humanoid robotics — from the Isaac GR00T foundation models that give robots their "brains," to Isaac Sim and Isaac Lab for training and testing robots safely in simulation before they're ever deployed in the real world, to its recent unveiling, alongside Unitree and Sharpa, of a complete open reference humanoid platform at GTC Taipei in June. Nvidia CEO Jensen Huang has repeatedly described physical AI — robots and machines that can sense, reason, and act in the real world — as a market opportunity worth potentially tens of trillions of dollars over the coming decade. But that vision depends entirely on solving a problem that has nothing to do with raw capability: convincing factory operators, hospital administrators, logistics companies, and ultimately regulators that these machines can share physical space with human workers without putting them at risk. That trust gap has been a persistent and explicitly acknowledged obstacle. As one industry report on humanoid robotics noted earlier this year, deployments of AI-powered humanoids have so far been largely limited to controlled environments like warehouses, with safety and privacy concerns continuing to pose real hurdles to broader use in less controlled settings, including eventually homes. Halos for Robotics is Nvidia's attempt to directly close that gap by giving the entire industry — not just Nvidia's own hardware customers — a standardized, independently accredited way to prove safety claims rather than simply asserting them. Part of a Broader Industrial Safety Push Halos for Robotics builds on related safety work Nvidia has been rolling out across its broader robotics ecosystem throughout 2026. At GTC earlier this year, Nvidia introduced its "Outside-In Safety" workflow, which supply chain company KION Group has adopted to enable what Nvidia calls outside-in perception — an approach where an AI agent strengthens a robot's own onboard safety mechanisms by drawing on infrastructure-mounted cameras and dynamic virtual safety fences positioned around a facility, rather than relying solely on a robot's own onboard sensors to detect nearby humans. Other partners are applying similar IGX Thor-based safety capabilities in even higher-stakes settings. Hexagon Robotics has adopted the platform for real-time AI reasoning and multimodal sensor fusion in its own humanoids, while in the surgical robotics space, companies including Johnson & Johnson, KARL STORZ, Medtronic, LEM Surgical, and Horizon Surgical Systems have all begun adopting or evaluating IGX Thor specifically for the kind of precision and functional safety that operating-room environments demand — a notably different but related category of physical AI safety challenge, where the margin for error is even smaller than on a warehouse floor. What It Means for the Humanoid Robotics Industry For an industry that has, until now, largely relied on individual companies developing and validating their own proprietary safety systems with limited standardization across the sector, Halos for Robotics represents a meaningful structural shift. By offering a shared hardware and software foundation alongside an accredited certification pathway, Nvidia is positioning itself not just as a chip and software vendor for humanoid robotics — a role it has aggressively pursued since at least CES 2026 — but as something closer to a de facto safety standards body for the entire physical AI sector, at least among the companies building on its platform. Whether that positioning translates into broad industry-wide adoption remains to be seen. Robotics companies building entirely proprietary, vertically integrated systems — Tesla's Optimus program is the most prominent example — may have less incentive to adopt an Nvidia-defined safety framework than companies like Agility, which already build extensively on Nvidia's broader hardware and software stack. But for the growing ecosystem of humanoid robotics companies that do rely on Nvidia's chips and software, Halos for Robotics offers something the industry has conspicuously lacked: an accredited, third-party-recognized way to answer the question that ultimately determines whether any of this technology gets deployed in spaces full of human workers — is it actually safe?