Physical AI at Scale Requires Safety Across the Entire Stack
With tens of millions of autonomous vehicles and industrial robots expected in the coming decade, NVIDIA argues that safety must be built into every layer of physical AI systems.
Physical AI is moving out of the lab and into the real world at scale. Industry analysts expect the installed base of level 3–5 autonomous vehicles to reach 49 million by 2035, according to ABI Research, while Omdia projects that about 60 million industrial robots will be deployed between 2026 and 2035. These systems will operate in unpredictable environments, making their safety requirements fundamentally different from those of traditional software.
NVIDIA argues that this shift demands a safety framework spanning every layer of the AI stack. Rather than treating safety as a single checkpoint or a post-training add-on, the company emphasizes continuous validation, monitoring, and control across the hardware, software, and operational layers. The goal is to ensure that physical AI can be trusted to interact with people and infrastructure at scale.
The source does not provide implementation details or compare approaches from other companies. It focuses on the necessity of layered safety as a precondition for mass deployment, driven by the projected growth in autonomous vehicles and industrial robotics.
More in AI & ML
Jev Creator on System One Models for Production, Not AGI
TypeSafe AI CEO Diogo Almeida, lead creator of Jev, argues System One models belong in production rather than on an AGI pedestal.
Meta's Muse AI Assistant Has a 0-Day That Lets Attackers Hijack It
A newly reported vulnerability in Meta's Muse AI assistant can be exploited with a simple ClickFix attack to take full control of the agent.
Pruning LLMs by Removing Blocks as an Ising Optimization Problem
A new approach frames large language model pruning as a physics-style Ising optimization to decide which blocks to remove.
NVIDIA: AI Security Needs Engineering, Not Just Policies
NVIDIA argues that securing AI agents requires treating security as an engineering discipline with requirements, controls, owners, and evidence.