The Data Transfer
When Pokémon Go launched in 2016, it became one of the fastest-growing mobile applications in history. The game required players to navigate their physical environments through their phone cameras, scanning the world around them in real time as they searched for virtual characters superimposed on live video feeds. What players experienced as a game mechanic was simultaneously a structured data collection exercise. Every scan captured three-dimensional spatial information about the surrounding environment: the angular geometry of buildings, the textures of surfaces, the dimensional relationships between physical objects in a given space.
According to reporting by Dutch newspaper Trouw, Niantic transferred approximately 30 billion of these three-dimensional scans to an American military contractor. Niantic has denied the accusations. Denial is, at this stage, the only commercially available response. The scale of what was reportedly transferred — across hundreds of cities, residential districts, transport hubs, and government areas in dozens of countries — represents a geospatial intelligence resource that no state actor could have assembled through any conventional means. The coverage of the dataset is a direct function of where the game was popular, which is to say: everywhere people carried smartphones.
What the Scans Train
The military application of the dataset addresses one of the central vulnerabilities of contemporary autonomous weapons systems: GPS dependency. Satellite positioning can be jammed. In Ukraine, electronic warfare along the front line has made GPS an unreliable navigation reference for both drones and munitions. Russian and Ukrainian forces have invested heavily in GPS spoofing and jamming capabilities precisely because so many battlefield systems — from loitering munitions to first-person-view drones — rely on satellite signals to know where they are and what they are targeting.
A neural network trained on 30 billion three-dimensional scans of real-world environments can learn to navigate and acquire targets through visual-spatial recognition alone — without any reference to satellite positioning whatsoever. The drone recognises its surroundings the same way a human would: by matching what its sensors observe against a learned spatial model of the world. In a jammed environment where GPS-dependent systems have been effectively blinded, a visually-navigating autonomous system remains fully operational. The Pokémon Go dataset, covering inhabited environments across the globe at a level of three-dimensional detail that no dedicated military survey programme could have generated without decades of effort and political access that does not exist, provides precisely the training corpus for that capability.
Millions of users, acting voluntarily and entirely without compensation, collectively performed a mapping operation of strategic value — in city centres, apartment blocks, transport infrastructure, and sensitive sites across dozens of countries — that no intelligence agency in the world could have commissioned directly.
The Keyhole Lineage
The relationship between Niantic and the American defence and intelligence community is not a matter of inference from corporate structure. It runs through the career of the company's founder. Before building Niantic, John Hanke founded Keyhole — a company specialising in three-dimensional geospatial mapping and satellite imagery. Keyhole's development was directly funded by In-Q-Tel, the venture capital arm established by the CIA specifically to invest in technology companies developing capabilities of interest to the American intelligence community. The funding relationship between Keyhole and In-Q-Tel is public record.
Google acquired Keyhole in 2004. Its core technology was subsequently incorporated into what became Google Earth — the satellite mapping platform now used by governments, militaries, and civilian users worldwide. The institutional chain runs from CIA-backed geospatial research, through one of the world's most widely used mapping tools, to the augmented reality gaming platform that generated 30 billion terrain scans now being used to train military drone navigation systems. Each step is commercially comprehensible. The cumulative trajectory is not accidental.
The Broader Pattern
The Pokémon Go case is not an anomaly in the relationship between consumer technology and defence applications. It is an example of a method. The logic is straightforward: data that would be prohibitively expensive, legally complex, or politically impossible to collect through declared intelligence or military programmes can be obtained at scale if users generate it voluntarily while believing they are doing something else entirely. The smartphone accelerated this dynamic by embedding persistent sensor arrays — cameras, microphones, gyroscopes, GPS receivers, accelerometers — into devices carried by billions of people into virtually every environment on earth.
What changed with Pokémon Go is the three-dimensional specificity of the data and the institutional directness of the path from consumer product to weapons development. Players searching for virtual characters were simultaneously performing detailed spatial surveys of their environments at a resolution and geographic breadth that no military mapping programme in history has matched. The game required no deception about what it was doing technically. The deception, if it can be called that, was simply about purpose.
Note: This intelligence note was completed on 11 June 2026. The core reporting on Niantic's data transfer draws on Trouw (Netherlands). Niantic's denial of the allegations is on record. In-Q-Tel's funding of Keyhole and Google's subsequent acquisition are matters of public record. This note does not constitute legal, investment, or policy advice.
