Pokémon Go Scans Trained Military Drone Navigation AI

From PokéStops to Battlefield: The Unseen Pipeline

A massive dataset, gathered unwittingly by millions of players seeking in-game rewards, has reportedly become a cornerstone for developing navigation technology destined for military drones and robots. According to reports from Dutch newspaper Trouw and subsequent coverage, roughly 30 billion environmental scans collected through Pokémon Go's scanning feature have been used to train a Visual Positioning System (VPS) model.

This technology is now at the heart of a December 2025 partnership between Niantic Spatial—the spun-off mapping and AI division of Niantic—and Vantor, the defense and intelligence contractor formerly known as Maxar Intelligence. The stated goal is to create a GPS-independent navigation system for drones and other assets operating in electronic warfare environments.

The core ethical dilemma is stark: data collected under the guise of gameplay has potentially been repurposed for military applications. While Niantic Spatial claims the Pokémon Go data is not part of its current Vantor agreement and Vantor denies using the game's data, experts argue the model's development was fundamentally accelerated by the player-generated scans.

The Scanning Mechanism: Consent Buried in Terms

Since 2021, Pokémon Go has encouraged players to record short, 360-degree videos of real-world PokéStops to earn extra items. This feature was optional, but granting permission meant agreeing to separate terms that gave Niantic a transferable, sublicensable license to the footage.

Most players, like Dutch player Floris De Hingh, saw it as a simple game mechanic. "I was just playing a game," he told Trouw, having even scanned his own apartment. The sheer scale of this data collection—billions of scans of streets, parks, and buildings—provided an unprecedented, global dataset for training AI to understand and navigate the physical world visually.

Niantic Spatial's CTO, Brian McClendon, a former lead for Google Maps and Earth, has stated this VPS approach is ideal for robots operating where GPS fails, whether in dense urban canyons or deliberately jammed war zones. The system works by matching a camera's live feed against a pre-built 3D model, needing only a few recognizable reference points to establish location.

The Defense Partnership: Fusing Ground and Air Maps

The partnership announced on December 16, 2025, aims to merge two systems. Niantic Spatial's VPS handles ground-level localization. Vantor's Raptor software, launched in February 2025, performs a similar function from the air using drone cameras and proprietary 3D terrain data.

Combined, they promise to allow a drone overhead and a ground unit below to share precise coordinates in real-time without any satellite link. This directly addresses a critical vulnerability exposed in modern conflicts: the ease with which GPS signals can be jammed or spoofed.

Vantor's own release bluntly targets GPS "unavailability, spoofing, interference, and jamming." The company, a prime contractor to the U.S. National Geospatial-Intelligence Agency, is no startup. Its rebranding from Maxar Intelligence in late 2025 underscores its deep roots in national security imagery and data.

A Trail of Ambiguous Denials

The companies' statements on the use of Pokémon Go data have been carefully worded and conflicting. When asked directly by Trouw, Vantor said it would not use the game's data but declined to say whether the model it plans to deploy was trained on those scans in the past.

Niantic Spatial, in statements to Kotaku, clarified that Pokémon Go data is not shared with them following their spin-off and Scopely's acquisition of the games business. They also stated ground scans were "one component" to train early versions of their AI models, but the final models are a product of that training, not a copy of the scans.

Jeroen van den Hoven, a professor of ethics and technology at TU Delft, told Trouw the conclusion is unavoidable: "Without the huge number of scans from all those gamers, the development of this system would never have progressed so quickly." He noted that once data is absorbed into an AI model, tracing its origins becomes nearly impossible.

Niantic's Geospatial Legacy and the Broader Data Landscape

The military link is less surprising given Niantic's lineage. The company grew from Keyhole, a mapping firm that received funding in 2003 from In-Q-Tel, the CIA's venture arm. Keyhole's technology was used to support U.S. troops during the Iraq War before Google acquired it.

John Hanke, Keyhole's CEO who later founded Niantic, led Google's geo teams. Niantic's first game, Ingress, also collected player-scanned imagery. The 2025 corporate split saw the games business sold to Scopely (backed by Saudi Arabia's sovereign wealth fund) while the mapping technology became Niantic Spatial.

This episode highlights a broader data ethics crisis. As Iris Muis of Utrecht University's Data School noted, users cannot foresee how their data might be used years later. Pokémon Go is not alone; Meta's smart glasses, Apple's AR hardware, and autonomous vehicle fleets continuously scan environments, building detailed maps with unclear future applications.

The Tangible Need for GPS-Denied Navigation

The technology itself solves a genuine and urgent military problem. DroneXL's reporting has extensively covered how GPS jamming has rendered many systems useless in conflicts like Ukraine, spurring rapid innovation in alternative navigation like terrain matching.

Ukrainian manufacturers, as noted in a Business Insider report, have developed drones at breakneck speed, tested rigorously by combat pilots like "Spring" to ensure they function in contested electronic environments. The U.S. Pentagon's Drone Dominance program and companies like Shield AI are also deeply invested in solving the GPS-denied challenge.

The ethical discomfort, therefore, is not with the goal of creating robust navigation. It lies in the sourcing of the training data from a global pool of gamers who consented to terms for a game, not a defense program. The line between consumer entertainment and military technology has blurred in a way that challenges fundamental notions of informed consent.

As field testing of the Niantic-Vantor system begins, the operational capabilities will become clear. The provenance of the data inside the model, however, will likely remain obscured, a baked-in ambiguity that is itself a powerful statement on the era of data-driven AI.