This week, technology observers took note of an unexpected convergence: the game that once sent millions of people wandering city streets chasing cartoon creatures is now teaching robots how to find their way around those same streets. Niantic Spatial has just teamed up with Coco Robotics, a startup that deploys last-mile delivery robots in a number of cities across the US and Europe.
The problem the partnership solves is real enough. In dense urban environments ringed by tall buildings, GPS signals bounce unpredictably. Niantic Spatial originally developed its visual positioning system for augmented reality applications, intending it as a method for anchoring digital content to the real world when users are looking at a specific location. It turns out the same technology works equally well for robots trying to understand their surroundings without relying on wobbly satellite data.
Niantic Spatial's technology will allow Coco's robots to position themselves in the correct pickup spots outside restaurants, making sure they don't get in anybody's way, and stop just outside the customer's door instead of a few steps away. Precision matters in last-mile logistics; every missed mark translates to inefficiency and frustrated customers waiting for their orders.
What makes this partnership significant is the dataset behind it. Niantic's AI spinout is training a new world model using 30 billion images of urban landmarks crowdsourced from players. These images did not come from corporate camera fleets or expensive mapping vehicles. They were taken by millions of people playing a mobile game, pointing their phones at buildings, statues, and street corners as they hunted virtual creatures.
The scale is staggering. Five hundred million people installed that app in 60 days, according to Brian McClendon, CTO at Niantic Spatial. Even now, the game still drew more than 100 million players in 2024, eight years after it launched. This long tail of engagement means Niantic has collected images from multiple angles, in different seasons, under varied lighting and weather conditions. Robots training on such diverse data perform more reliably in the messy reality of urban delivery.
Here is where the tensions emerge. When Pokémon Go first launched in 2016, players were not informed that their casual photos of landmarks would one day power commercial robot navigation systems. The app added what it called "Field Research" in 2020, a feature prompting players to scan real-world statues and landmarks with their cameras in exchange for in-game rewards. These scans were presented as a feature that deepened gameplay. Their true function was data harvesting.
This raises legitimate questions about data governance that reasonable people disagree on. Niantic's position is legally defensible. The company retains those rights because that is precisely what players agreed to when they installed the game; Section 5.2 of the Niantic Terms of Service, titled "Rights Granted by You – AR Content", states that the company retains wide-ranging rights over anything that users upload. Legally, the company has done nothing wrong. The terms are clear, even if few people read them.
Yet a gap exists between legal permission and public expectation. Most players understood themselves to be helping build a game. Few likely understood they were contributing to a dataset that would eventually train robots competing for sidewalk space in their cities. This is a stark example of how crowdsourced data, seemingly collected for one purpose, can be quietly repurposed years later for something quite different.
The counterargument deserves serious consideration. Robots may need to share spaces with humans such as construction sites and sidewalks, and if robots are ever going to assimilate into that environment in a way that's not disruptive for human beings, they're going to have to have a similar level of spatial understanding. Better robot navigation plausibly makes shared urban spaces safer. If Niantic's technology helps delivery robots avoid collisions or move more efficiently, the public benefit is real.
Coco Robotics deploys around 1,000 flight-case-size robots built to carry up to eight extra-large pizzas or four grocery bags in Los Angeles, Chicago, Jersey City, Miami, and Helsinki; according to CEO Zach Rash, the robots have made more than half a million deliveries to date, covering a few million miles in all weather conditions. These machines are not hypothetical. They are operating now, and they benefit from better navigation technology.
The deeper issue is transparency. A company has the legal right to repurpose data in ways users did not anticipate. But there remains a case for telling users what is happening and why. Niantic's Privacy Policy would seem to indicate that uploaded AR imagery is anonymised during processing, and as such doesn't need to be treated in the same way that personally identifiable information would be; players have the right to opt-out of uploading additional data going forward, but can't remove what's already been pushed into the system. That one-way valve is worth questioning.
Technology companies will continue to find novel uses for data they already hold. That pattern is not unique to Niantic; it reflects how digital markets work. The question for users and regulators is not whether this will happen, but whether companies should be required to communicate these shifts more clearly. It is possible to believe both that Niantic acted within its rights and that greater transparency would have been fairer to the millions of people who contributed to this dataset without full knowledge of its eventual use.