Worldmodeldata Turns Licensed Gameplay Into Training Data for Physical AI Models

Worldmodeldata, a Cambridge-based startup, is pitching video games as a structured data source for AI systems that need to understand action, motion and consequence in the physical world. The company has raised GBP 7 million in seed funding and says it is building toward a library of 1 million hours of gameplay-derived training data by the end of 2026.