AI news · October 9, 2026

Nvidia's Halos adds a safety system for robots working near people

roboticsphysical-aisafetynvidia

Nvidia's Halos for Robotics combines safety compute, sensor links, software, simulation, and inspection; Ars says Agility is using it for Digit, making safety engineering part of the robot product.

Made With Models illustration for this story

An Oct8 Ars update examined Nvidia Halos for Robotics, a safety system first announced in June. It combines IGX Thor compute, Holoscan Sensor Bridge, Halos OS, simulation, and an inspection lab for robots such as warehouse vehicles and humanoids. Nvidia says Agility is integrating it into Digit, and more than 40 companies are involved in the inspection program.

The primary release says Halos Core is in early access for registered developers and the open-source Outside-In Safety Blueprint is also early access. This is not a general robotics SDK or a guarantee that a robot is safe. It is a concrete pattern for products that sense, decide, and move: monitor failures, isolate safety work, test in simulation, then seek independent certification.

What you can do with it

If you build a robot or sensor product, read the Halos release and map its compute, sensor, simulation, and certification pieces to your own design. The public blueprint is early access, so use it for a lab prototype and keep a human stop path during every physical test.

Our take

The valuable change is treating safety as part of the product stack instead of a final checklist. Nvidia has a strong incentive to frame its own system as the answer, and early access is not proof of real-world performance. The useful takeaway is the sequence: sense, isolate, simulate, inspect, then deploy.

Source: Ars Technica ↗ — Made With Models writes the brief; the reporting is theirs.