On June 22, 2026, at the Automate show in Chicago, NVIDIA announced "NVIDIA Halos for Robotics," a safety system for physical AI and real-world robots. The company positions it as the industry's first full-stack safety system, integrating AI compute, functional safety, sensors, software and certification into a single architecture. (announcement )
June 22, 2026 · NVIDIA
Halos for Robotics: A Full-Stack Safety System for Physical AI
NVIDIA extends its autonomous-vehicle safety platform to industrial and humanoid robots — unifying hardware, OS, and an accredited inspection lab into one stack, with Agility Robotics' Digit as the first partner.
18,600+
engineer-years of AV safety experience carried over
40+
partners across software, chips, and applications
1st
ANAB-accredited AI systems inspection lab (ISO/IEC 17020)
The Four-Part Stack
IGX Thor / Holoscan Sensor Bridge
Industrial-grade AI compute and sensor connectivity with built-in safety
Halos OS / Halos Core
Software stack providing safety-related functions
Halos Inspection Lab
ANAB-accredited lab verifying functional, AI, and cyber safety
Outside-In Safety Blueprint
Open-source design monitoring robot fleets via external cameras
From Development to Deployment
A standardized architecture spanning the full robot safety lifecycle.
Development
Architecture + AI models
→
Inspection
Functional · AI · cyber
→
Certification
Third-party labs
→
Deployment
Factories · warehouses
Seen as a strength
Mature AV-derived safety credentials transplanted to robots
"Outside-In" external cameras supervise multiple robots at once
First Digit deployments aimed at Amazon, GXO, Schaeffler, Toyota
Open questions
Final third-party certification still required separately
AI model safety assurance and edge-case validation are next
Pricing and benchmarks undisclosed; early-access stage
The Outside-In Approach
Beyond robot-mounted sensors, infrastructure cameras monitor entire fleets with low-latency detection and dynamic safety zones — supervising multiple robots at once, while raising the question of whether safety holds if a single sensor fails.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…