NVIDIA pitches Halos as the first full-stack safety system for autonomous vehicles and robots
NVIDIA says its Halos system spans hardware, software, AI behavior, simulation and certification readiness, and argues that safety proof is what turns a physical AI prototype into a deployable product.
- 49 million
- 60 million
- more than a decade
What Happened
Physical AI is moving rapidly from research into large-scale deployment. ABI Research projects an installed base of 49 million Level 3-5 autonomous vehicles by 2035, while Omdia estimates that roughly 60 million industrial robots will be deployed between 2026 and 2035. As these machines enter roads, factories and warehouses shared with people, safety has to scale with them, NVIDIA argues. The company says proving that AI-driven machines behave safely when their decisions become physical action requires safety across hardware, software, AI, the operating environment and the deployment lifecycle, not a one-time check.
- Dynamic environments require context-aware safety: roads, factories and warehouses cannot be fully controlled with static zones or physical barriers, so systems must perceive changing conditions, adapt and reach a safe state when something unexpected happens.
- AI behavior requires its own assurance: testing must assess AI software alongside traditional functional safety using design-time, runtime and validation-time guardrails, and emerging standards such as ISO/IEC TS 22440 are beginning to address AI-specific risks.
- Deployment is ongoing: autonomous vehicles and robots evolve through software and model updates, new tasks and changing operating conditions, and material changes may require additional safety testing.
- Validation at scale requires simulation and synthetic data: the number and complexity of potential scenarios means real-world testing must be combined with simulation, synthetic data generation and scenario reconstruction.
NVIDIA says physical AI safety demands specialized engineering, data, processes and validation that few companies can reproduce alone. Its safety foundation draws on more than a decade of autonomous vehicle safety development, covering functional safety, sensor fusion, AI behavior assurance, vision AI, simulation and real-world validation. The company calls NVIDIA Halos the first and only full-stack safety system for physical AI, engineering safety across every layer of design, validation and deployment, with principles shared across vehicles and robotics while platforms, standards and evidence stay specific to each domain. The Halos AI Systems Inspection Lab turns safety, cybersecurity and AI safety requirements into repeatable inspections and helps prepare integrations for final system-level certification by third-party agencies.
- Hardware: NVIDIA DRIVE AGX Thor provides safety-engineered accelerated compute, while NVIDIA Hyperion supplies the full-stack vehicle platform and reference architecture for Level 4 autonomous vehicles.
- Operating system and middleware: Halos OS provides a unified software foundation built on ASIL-D certified DriveOS, with Halos Core and Halos Middleware supporting system isolation, monitoring and deterministic communication.
- End-to-end model: NVIDIA Alpamayo offers open reasoning vision language action models that bring explainability to long-tail scenarios.
- Simulation and validation: the NVIDIA Halos Safety Evaluation Framework provides tools and guidelines for generating evidence to support autonomous vehicle safety cases across different levels of automation.
- Together these connect cloud-based AI development and simulation with in-vehicle deployment so safety evidence stays traceable across the vehicle lifecycle.
- Hardware: NVIDIA IGX Thor is an industrial-grade module combining accelerated computing and functional safety on one platform with a dedicated Functional Safety Island, designed to support systems built for standards including IEC 61508 and ISO 13849.
- Software: Halos Core for IGX provides the foundation for safety-related operating functions including fault detection, monitoring and reporting, plus communication and processing that connects sensors, actuators and other safety components.
- Real-time sensing: NVIDIA Holoscan Sensor Bridge connects sensor data with AI and safety-related processing, helping systems identify invalid information and execute defined safety responses.
- Simulation and validation: NVIDIA Isaac Lab and NVIDIA Omniverse libraries let developers test robot behavior across relevant conditions and edge cases, complementing real-world validation.
- Outside-in safety: the open source NVIDIA Halos Outside-In Safety Blueprint uses external cameras and vision AI agents to extend awareness beyond onboard sensors for facility-level monitoring and functional safety.
- In autonomous vehicles, Geely, Isuzu, Nissan (powered by Wayve software) and Einride are building Level 4-ready vehicles on NVIDIA Hyperion, supported by Halos OS.
- Uber, Grab, Lyft and other mobility providers are using Hyperion to scale robotaxi development and deployment.
- Halos AI Systems Inspection Lab members include AUMOVIO, Bosch, Gatik, Hesai, Lucid, MIRA, onsemi, PlusAI, Sony, Valeo and Wayve, spanning autonomous driving, ADAS, sensors, silicon, systems integration, validation and safety assurance.
- In robotics, acontis and QNX provide embedded software for predictable safety functions, Advantech and NexCOBOT build safety-designed IGX systems, and Infineon, NXP, STMicroelectronics and Texas Instruments contribute sensor and safety-microcontroller technologies.
- KION Group is developing functional safety agents for autonomous forklifts, and Agility is integrating NVIDIA IGX Thor and Halos Core into the safety system for its Digit 5 humanoid.
Independent assessment is central to the pitch. For autonomous vehicles, TÜV SÜD certified NVIDIA's Automotive Product Lifecycle software process and DriveOS 6.0 to ISO 26262 ASIL D, and certified NVIDIA's automotive engineering processes to ISO/SAE 21434, while TÜV Rheinland performed an independent UNECE safety assessment of NVIDIA DRIVE AV. For robotics, TÜV Rheinland is inspecting NVIDIA IGX Thor, Halos OS and Holoscan Sensor Bridge for functional-safety certification readiness, building on TÜV SÜD's inspection of the Thor SoC and Halos Core for ISO 26262. Across physical AI, ANAB has accredited the NVIDIA Halos AI Systems Inspection Lab as an ISO/IEC 17020 inspection body, inspecting scoped Halos integrations and helping companies prepare for final certification by independent third parties.
Previously from NVIDIA Corporation
NVIDIA has been building out this safety layer for months. In June 2026 it introduced Halos OS as a certified safety foundation for robotaxis, aimed at regulatory requirements for reliable autonomous vehicle operation. In August 2026 it released Alpamayo 2 Super, described as the top-ranked open reasoning model for autonomous driving, commercially available under a permissive license. In September 2026 it detailed a three-computer robotaxi stack for training, simulation and in-vehicle computing that it says every major commercial robotaxi program runs on, in a market projected to reach $400 billion by 2035.
- NVIDIA Halos OS Provides Safety Foundation for Robotaxi Deployments
- NVIDIA Alpamayo 2 Super Open Model for Robotaxis and AVs Now Commercially Available
- NVIDIA details its three-computer robotaxi stack as partners target driverless fleets at scale
Background drawn from MotorClaw's earlier coverage of NVIDIA Corporation's official releases.
Why this matters
Manufacturers, regulators, insurers and workplace safety teams all need evidence that automated hardware, software and AI can work safely around people before autonomous vehicles and robots scale. NVIDIA says its Halos ecosystem already spans vehicle makers, mobility providers, chip suppliers and certification bodies, so its safety approach could shape how quickly both industries deploy.
Terms in This Story
- Physical AI
- AI-driven machines, such as autonomous vehicles and robots, whose decisions result in physical action in the real world.
- Functional safety
- An engineering discipline focused on making sure a system responds safely when something fails, rather than only working correctly when nothing does.
- ASIL D
- The highest automotive safety integrity level defined by the ISO 26262 functional safety standard for road vehicles, assigned to systems where failure carries the greatest risk.
- Robotaxi
- A self-driving vehicle operated as a commercial ride-hailing service without a human driver.
Summarised from the linked release; details can be imperfect — always verify against the original source.