LONDON, June 12 — The summer of 2016 gave us many things: an epidemic of people walking into lamp posts, strangers gathering in public parks at midnight, and the sight of office workers suddenly sprinting down city blocks because someone had spotted a Snorlax.
Pokémon Go wasn’t just a game. It was a global social experiment powered by curiosity, nostalgia and the irresistible urge to catch ‘em all.
Now, nearly a decade later, some of the data generated through those adventures has resurfaced in an entirely different context.
According to reporting by The Guardian, AI models trained using historical scans collected through Pokémon Go are connected to a partnership aimed at helping drones navigate environments where GPS signals are unavailable or compromised.
The story centres on Niantic, the company behind Pokémon Go. In 2021, the game introduced a feature that encouraged players to scan real-world locations known as PokéStops in exchange for in-game rewards. Participation was optional, requiring users to actively opt in and upload footage captured on their devices.
Those scans, Niantic said, helped train its AI models to better understand physical spaces.
Fast-forward to today, and the company has confirmed a partnership with Vantor, a firm specialising in spatial detection software for drones, including some used by military organisations.
Both companies told The Guardian that Pokémon Go scans themselves were not handed over to Vantor. Instead, the scans had been used to train Niantic’s broader foundation models.
It’s a distinction that matters technically, though perhaps less so emotionally for players discovering that their quest to evolve a Magikarp may have contributed, however indirectly, to technology with potential defence applications.
The revelation taps into a growing discomfort around the afterlife of data. Information collected for one purpose can eventually find itself serving another.
Tom Sulston of Digital Rights Watch argued that while users may technically agree to terms and conditions, most people aren’t combing through lengthy legal documents before downloading a free game.
He suggested regulators should place greater emphasis on what serves the “best interests” of users and whether certain practices are “fair and reasonable”, according to The Guardian.
Researchers say this may only be the beginning. Dr Rob Nicholls, associated with the University of Sydney’s Centre for AI, Trust and Governance, noted that fitness tracking data has previously been used to identify sensitive military locations, highlighting how seemingly ordinary technologies can reveal unexpected consequences.
The broader lesson isn’t necessarily that augmented reality games are dangerous. Rather, it’s that data has a remarkably long shelf life.
The digital breadcrumbs we leave behind — our routes, routines and interactions with the world around us — can outlast the moment in which they were created.
For millions of Pokémon Go players, scanning a local mural or playground probably felt like just another side quest.
Years later, those same actions have become part of a much bigger conversation about consent, transparency and who ultimately benefits from the data we generate.
In the age of AI, it turns out that even catching Pikachu can have consequences no one saw coming.
