Pokémon Go Players Unknowingly Built the Map Guiding Military Drones

Roughly 30 billion environmental scans captured by Pokémon Go players since 2021 — short videos of streets, buildings, and parks recorded for in-game rewards — became the training data for a Visual Positioning System now headed into U.S. military drones. The scans are owned by Niantic Spatial, which in December 2025 partnered with defense contractor Vantor (formerly Maxar Intelligence, a $70M prime contractor to the National Geospatial-Intelligence Agency) to fuse its ground-level camera navigation with Vantor’s aerial Raptor software. The combined system lets drones and ground units share coordinates in GPS-denied environments, a capability in heavy demand as jamming and spoofing spread from battlefields like Ukraine into civilian airspace. Field testing is planned for early 2026.

The companies’ answers on data provenance are evasive. Vantor told Dutch paper Trouw it ‘would not use’ the game’s data but declined to say whether the model was trained on it historically; Niantic Spatial has acknowledged the scans trained an ‘early version’ of its navigation model. TU Delft ethicist Jeroen van den Hoven argues the system’s rapid development would have been impossible without the player scans — and that once data is absorbed into a model, proving its presence or absence becomes practically impossible. Players agreed to transferable, sublicensable license terms, but none were told their footage might steer weapons platforms; one early player noted he had scanned the inside of his own apartment.

The trajectory is less surprising given Niantic’s lineage: it descends from Keyhole, a CIA-backed (In-Q-Tel) mapping firm whose tech supported U.S. troops in Iraq before Google acquired it. In 2025 the company split — its games sold to Saudi-owned Scopely for $3.5B while the mapping platform became Niantic Spatial under founder John Hanke. The episode crystallizes a broader consent problem: Meta’s smart glasses, Apple’s AR hardware, and Waymo’s vehicles are all building 3D world models, and users have no way to anticipate how that data will be repurposed years later.