Nvidia expands Halos safety system to robotics
Nvidia announced Halos for Robotics in June 2026, extending its safety architecture from autonomous vehicles to robots. The company says the system is designed to help reduce the risk of machines harming nearby people.

Key takeaways · 5
- 01
Nvidia’s Halos for Robotics is intended for warehouse mobile robots, factory humanoids and surgical robots.
- 02
IGX Thor includes an independent processor for safety workloads, while Halos isolates safety-critical computing to avoid potential interference.
- 03
Agility expects early access to Digit 5 in the first half of 2027 and general availability by the end of 2027.
- 04
Nvidia’s inspection lab provides integration feedback and helps prepare work for independent certification; it does not automatically certify finished robots.
- 05
Agility reported more than $300 million in multi-year Digit 5 customer orders as of May 2026, contingent on contractual milestones.
A safety framework for different robots
Nvidia first launched Halos in 2025 as hardware and software tools for autonomous-vehicle safety.[1] Its June 2026 announcement extends the architecture to robotics under the name Halos for Robotics.[1] Nvidia describes Halos as a full-stack safety system that companies can build on to reduce the risk of machines harming nearby people.[1] The intended applications include warehouse mobile robots, factory humanoids and surgical robots.[1] Nvidia has also said safety considerations vary by application, so the system is not presented as a single, identical safety configuration for every robot.[1] An Nvidia speaker characterized safety as a possible next bottleneck as AI models and robot hardware become more capable.[1]
Computing, sensors and developer controls
Nvidia’s IGX Thor computing module includes an independent processor for safety-related workloads, and Halos isolates safety-critical computing to avoid potential interference.[1] The Holoscan Sensor Bridge connects sensor data to safety-related computing and can identify corrupted data.[1] The package also includes simulations in virtual environments and an inspection lab that gives partners quick feedback.[1] Nvidia says the lab is intended to help prepare integrations for final certification by independent third-party bodies, rather than automatically certify finished robots.[4] Developers get hooks to define custom safety functions.[1] These features offer building blocks for integration, but the evidence does not establish that they produce lower real-world injury or crash rates.[3]
Digit 5 is the first integration
Agility Robotics was the first company to incorporate Halos into a robot, pairing its Digit 5 humanoid with Nvidia IGX Thor and Halos Core alongside Agility’s own safety systems.[1][2] Agility says Digit 5 was engineered to work near people without the physical safety barriers used in traditional automation.[2] The robot can autonomously avoid, stop or assume a seated position.[2] Agility says the design builds on more than 65,000 hours of real-world robot operation.[2] It also reported more than $300 million in multi-year Digit 5 customer orders as of May 2026, contingent on contractual milestones.[2] That order figure is not a measure of delivered robots or completed deployments.[2]
Availability and independent evaluation
Agility expects early access to Digit 5 in the first half of 2027 and general availability by the end of that year.[2] Those dates are plans, not confirmation that the robot is already broadly available; a report noted the dates could slip and that Digit 4, not Digit 5, was the Agility robot operating at work at the time of that report.[3] Agility said Digit 5 passed an independent field evaluation on a customer production line for industrial safety standards administered by OSHA.[2] Nvidia named Boston Dynamics an early partner in its Halos safety accreditation program, covering Spot, Stretch and Atlas, and said KION Group and LG were also working to incorporate Halos.[1] The evaluation and partner activity do not establish lower real-world injury or crash rates.[3]
For teams evaluating robots that work near people, the announcement offers a defined set of safety-oriented computing and integration tools to assess alongside application-specific requirements. It also makes the distinction between engineering support, independent certification and demonstrated real-world safety outcomes important when planning a deployment.
Why it matters
Put this to work — one session a day, built for your industry.
Create a free account for a daily session — eight questions and one real-work challenge, on the news that affects your role.
Start freeHow this developed
21 October 2026
Nvidia expands Halos safety system to robotics
Sources
- Nvidia's big bet on physical AI aims for safer robotaxis, humanoid robots - Ars Technicaarstechnica.com
- Agility Unveils Digit 5 Humanoid Robot Built for Cooperatively Safe Work at Scale | Agilityagilityrobotics.com
- Nvidia's safety assessments don't establish safer machinesvector.news
- Why Physical AI Needs Safety at Every Layer | NVIDIA Blogblogs.nvidia.com