LONDON — Google DeepMind began integrating Street View data into its Project Genie world model on May 19, 2026, with the feature unveiled during the Google I/O developer conference. Access opened the same day to Google AI Ultra subscribers in the United States, with Ultra users outside the U.S. set to gain access over the following weeks.

Project Genie is a general-purpose world model that generates interactive environments, with applications spanning educational experiences, gaming, and robotics training. Genie 3, which Google opened to U.S. AI Ultra subscribers in January 2026, lets customers create interactive game worlds from text prompts or images.

The Street View integration enables simulations anchored to real-world locations and allows the point of view to shift from cars to other agents such as humans or robots. Google has collected Street View data for 20 years using camera-equipped cars and individuals wearing tracker backpacks, amassing more than 280 billion images across 110 countries and seven continents.

"We have long thought about how we can build out the best and richest model of the world on top of Street View data," Jonathan Herbert, director of Google Maps, said. Herbert started on the Street View team as an intern 12 years ago. "It's definitely been an idea of ours to use Maps Data in new ways and for new kinds of AI research for a pretty long time," he said.

Jack Parker-Holder, a research scientist on DeepMind's open-endedness team, described use cases ranging from robotics training to consumer exploration. Genie can simulate occasions when the sun glints off Victorian housing in London so that a robot does not get shocked by the rays. "It's really powerful for both the agent [and robotics] use case and for humans to play with, and that's always been the thesis of Genie," Parker-Holder said. "Simultaneously, you might say, 'I'm going to New York City, but not this time of year. It's going to be snowy. I want to see what that block looks like in the snow,'" he added.

Genie 3 also powers one of Waymo's simulators, which trains self-driving cars on scenarios such as tornadoes or elephant encounters. Waymo has used a simulator to scale to 11 U.S. cities and test its AI driver in several more, and could use Street View in Genie to prepare launches in additional cities globally.

Diego Rivas, a product manager at DeepMind, said that Street View and Genie are experiments that require improvements in accuracy. Sample simulations, including an underwater rendering of a neighborhood, produced recognizable but video game–quality results rather than photorealistic ones. The models are not yet physics-aware and do not understand cause and effect; in one simulation, a woman running through a snowy Joshua Tree ran through cacti and bushes. Physics is not hard-coded into the models, which learn it over time through passive observation. "I think for this kind of model, it's maybe six to 12 months behind video in terms of the accuracy and quality, so I think it's something we will solve," Parker-Holder said.