Nvidia has unveiled Halos for Robotics, which it calls the industry's first full-stack safety system for "physical AI" — the term for artificial intelligence that controls machines moving through the real world, like humanoid robots.
The pitch is straightforward. According to Bloomberg's Ian King, Nvidia argues that humanoid robots will need to handle split-second decisions before they can be trusted to work closely with people. Halos is meant to provide that safety layer.
Techmeme and Bloomberg report that Halos is a safety-focused operating system derived from Nvidia's autonomous-vehicle technology, designed to run on the company's IGX Thor hardware. The carryover makes sense: the problem of a self-driving car avoiding pedestrians is closely related to a factory robot avoiding workers.
Nvidia is also opening a lab. According to GlobeNewswire, the NVIDIA Halos AI Systems Inspection Lab is the world's first program accredited by the ANSI National Accreditation Board (ANAB) for functional and AI safety in physical AI, intended to help partners prepare their systems.
Early partners are already on board. SiliconANGLE and Investing.com report that Agility Robotics became the first company to use Halos to build safety into robots working in factories and warehouses. Separately, FORT Robotics announced it is extending its "Trust Layer for Physical AI" by adding what it calls Outside-In Safety in collaboration with Halos.
Why it matters: as humanoid robots move from demos into real factories and warehouses where they share space with people, standardized safety systems and independent certification could determine how quickly — and how safely — these machines get deployed alongside human workers.