How Pokémon Go Data Trained AI Now Powering Military Drones

The massive datasets generated by millions of Pokémon Go players are finding a high-stakes new application in defense technology. Through Niantic’s spatial AI models, voluntary augmented reality (AR) scans are being repurposed to enable autonomous drones to navigate without the need for GPS.

From Mobile Games to GPS-Denied Navigation

The transition from gaming to defense began with a 2021 update to Pokémon Go, where Niantic introduced in-game incentives for players to scan real-world environments. This movement turned millions of users into voluntary data collectors, generating billions of visual mapping data points encompassing streets, buildings, parks, and trees.

This massive repository of ground-level visual data was used to train Niantic's foundation models. Following the spin-off of Niantic Spatial, this technology has moved beyond entertainment. The company has entered a strategic partnership with Vantor, a US defense intelligence firm, to integrate Niantic’s Visual Positioning System (VPS) with Vantor’s Raptor software.

Solving the Problem of GPS Jamming and Spoofing

In modern conflict zones, such as those seen in Ukraine and Iran, GPS jamming and spoofing are frequently used to disorient kamikaze drones, reconnaissance UAVs, and GPS-guided missiles. When satellite signals are compromised, autonomous systems typically lose their ability to navigate accurately.

The combination of Niantic Spatial’s ground-level 3D maps and Vantor’s 20 years of satellite imagery creates a shared coordinate system that allows machines to locate themselves using only visual input. This "GPS-free" navigation is immune to standard signal jammers. Early technical tests reported impressive results, including an error reduction of up to 70% and a navigational accuracy of approximately 1.5 meters.

The Commercial and Defense Landscape

The implications of this technology extend into large-scale military contracts. In February 2026, Vantor secured a US Army contract valued at up to $217 million for the One World Terrain program. This program aims to build high-precision 3D terrain data for the Army's Synthetic Training Environment, which is used for realistic mission rehearsals and simulations.

While Niantic has clarified that raw player scans were not handed directly to defense contractors—noting that the data was used to train general foundation models via opt-in user permissions—the lineage of the technology is clear. The corporate structure has also shifted to reflect this specialization: Niantic’s gaming division was acquired by Scopely for $3.5 billion in March 2025, while Niantic Spatial operates as a standalone entity focused entirely on geospatial AI.

Why This Matters for the AI Industry

This development highlights a growing trend in AI: the "dual-use" nature of massive consumer datasets. Data collected for low-stakes consumer engagement (like catching digital monsters) is becoming the bedrock for high-stakes industrial and military autonomy. As spatial AI matures, the ability to turn visual pixels into precise, navigable 3D coordinates will be the defining factor in the next generation of autonomous robotics.

Key Takeaways