19 March 2026 
The Danish cobot (collaborative robot) manufacturer Universal Robots (UR) has teamed up with Scale AI – a US company that supplies high-quality data for AI models – to develop a robot teaching system that uses data collected from robots working in real applications. They say that their AI Trainer technology marks “a tectonic shift”, allowing robots to move from pre-programmed applications to fully AI-driven tasks, bridging the current gap between laboratories and factories.
“Our customers, ranging from large enterprises to AI research labs, are no longer just asking for AI features,” says Anders Beck, vice-president of AI robotics products at Universal Robots. “They need a way to collect high-fidelity, synchronised robot and vision data to train AI models on the same robots they intend to deploy. Our AI Trainer is the industry’s first direct lab-to-factory solution for AI model training.”
Most of the data used to train robotics AI today is collected from research robots not designed for production settings, and many systems rely only on visual feedback, making delicate or contact-rich tasks difficult. Although these systems are effective for training physical AI models, they lack the power and reliability for long-term real-world deployment, according to UR.
“The AI Trainer directly addresses these barriers,” says Beck. “By using our unique Direct Torque Control and force feedback features, we give developers direct influence over how the robot physically interacts with the world, training on the same robust hardware used in over 100,000 industrial deployments.”
The AI Trainer lets human operators guide UR robots through tasks in a leader-follower set-up, automatically capturing high-quality data for robotics AI development. The operators guide a “leader” robot through a task, while a synchronised “follower” robot mirrors the motion in real time. The system records synchronised motion, force and visual data, producing the data needed to train VLA (vision-language-action) models
“Universal Robots is a leader in industrial robotics, and its global footprint offers the ideal foundation for data capture and AI deployment,” comments Ben Levin, general manager of physical AI at Scale AI. “Together, we’ve created an integrated robotics data flywheel, allowing customers to train, deploy, and improve their AI models faster than ever before.”
“Scale shares our belief that the future of robotics AI will be shaped as much by data quality as by model architecture,” adds James Davidson, chief AI officer at Teradyne Robotics, which owns UR and its sister company, MiR. “Together, we’re creating a foundation where imitation learning can move beyond isolated research projects and become a scalable, industrial capability. That’s an important step for customers who want AI systems they can actually deploy and rely on.”
The AI Trainer technology had its first public demonstration at Nvidia’s GTC 2026 AI showcase and conference in California this week. Attendees were shown two UR3e “leader” robots providing haptic input to control two UR7e “follower” robots.
Later this year, Scale and UR plan to release a large-scale industrial robotics dataset that captures economically useful work across a wide range of industrial and commercial environments. This will serve as robust pre-training data for customers who will train on UR’s robots. Scale’s software will be embedded into the UR AI Trainer for on-the-edge capture of fine-tuning data and real-world rollout data to improve models over time.
Leave a comment