Uber and Wayve launch supervised robotaxis on London streets

Wayve and Uber turning on supervised robotaxis in central London is a milestone for AI systems that must continuously perceive, reason and act in the messiest possible real-world environment. Compared to highly mapped geofenced deployments, London’s streets are dense, chaotic and full of edge cases. Successfully operating there, even with a safety driver, is a strong signal that end-to-end learned driving stacks are maturing beyond controlled pilots.

Strategically, this launch is important because it shows an AI-first AV company pairing with a global network orchestrator instead of trying to own the full stack alone. Wayve brings the model and driving brain, Uber brings demand, routing and regulatory muscle. If this model works, it may become a template for how AI driving companies scale internationally without burning billions on their own fleets and rider apps.

For the race to AGI, the deployment does not directly change model capabilities, but it validates a key ingredient: agentic systems that close the loop between perception, planning and long-horizon decision making under safety constraints. Techniques proven here will bleed into other domains where AI agents must operate in open-ended physical or digital environments, from warehouse robotics to software agents acting on desktops.