Nvidia has introduced Halos, what it calls the industry’s first full-stack safety system for humanoid robots, drawing on more than 18,600 engineering years of autonomous vehicle development. The architecture integrates AI compute, safety software, sensor data and an accredited inspection lab to enable robots to make split-second decisions and operate alongside people without halting. In parallel, San Francisco-based startup Robot.com—formerly Kiwibot—has begun commercial deployments of R-noid, a wheeled humanoid designed for repetitive tasks in food service, logistics and light industry. Fewer than 40 units are currently active across about a dozen customers, including a New York golf course where the robot loads delivery bots and packs orders.

Halos provides a unified safety layer: IGX Thor hardware processes real-time sensor feeds, while the Halos OS and an Outside-In Safety Blueprint use external cameras and AI agents to adjust robot behaviour in dynamic settings. Robot.com’s R-noid relies on custom foundation models developed with AI lab Physical Intelligence, targeting roughly 70% autonomy during initial deployment with remote teleoperation support. The company says it can fine-tune a robot’s brain with as little as 50 hours of task-specific data and have a unit operational within eight to twelve weeks.

Agility Robotics is the first adopter of Halos for its Digit humanoid, already deployed in Toyota manufacturing plants in Canada. Barclays estimates the humanoid robotics market could generate $200 billion in revenue by 2035. While physical safety systems advance, separate research from the University of Duisburg-Essen and the Technical University of Berlin examines the emotional dimension of human-AI interaction. A study published in Computers in Human Behavior: Artificial Humans, based on written interviews with 30 adult users of the Replika chatbot, found that half reported developing deep emotional relationships, with some describing physical sensations akin to romantic attraction. Researchers note that users often treat the AI as a non-judgemental companion, similar to a pet, but warn that these bonds are entirely dependent on corporate infrastructure—if a company shuts down, users lose the entity they are attached to.

Nvidia has opened an ANSI National Accreditation Board-accredited inspection lab to help robotics firms prepare for third-party safety certification from bodies such as TÜV Rheinland and UL Solutions. In Germany, researchers are advocating for legislative frameworks to govern emotional AI, emphasising the need for ethical oversight, data protection and safeguards for minors without resorting to outright bans. The next milestones to watch are the first Halos-certified robot deployments and any regulatory moves in Europe to address AI companionship.