Pokémon Go scans trained AI now powering military drones

Key Takeaways

- 30 billion AR scans submitted by Pokémon Go players trained Niantic Spatial's foundation models for GPS-free navigation
- Niantic Spatial partnered with defense contractor Vantor in December 2025 to enable drones to operate when satellite signals are jammed
- Vantor separately landed a $217 million US Army contract for 3D terrain data used in military simulations
Millions of Pokémon Go players spent years scanning real-world locations for in-game rewards. Those 30 billion AR scans trained AI models now being used by a US defense contractor to keep military drones flying when GPS fails. The connection between a mobile game and autonomous weapons systems traces back to a 2021 update that few players likely considered at the time.
Niantic Spatial, a spin-off from the game's developer, announced a partnership with defense intelligence firm Vantor in December 2025. Their combined technology addresses a critical battlefield problem: GPS jamming and spoofing already throw off kamikaze drones, reconnaissance systems, and guided missiles in Ukraine and Iran. When satellite signals go dark, troops and autonomous systems lose their bearings.
How did Pokémon Go scans end up in defense AI?
In 2021, Niantic added in-game incentives for players to scan real-world locations with their smartphone cameras. Parks, streets, buildings, trees. Participation was opt-in, with permissions collected, and covered by the privacy policies in place at the time. According to DroneXL, players generated billions of visual mapping data points.
Niantic Spatial transformed those scans into a large-scale 3D map that works without satellite signals. Machines equipped with cameras can locate themselves and navigate using visual references alone. The company calls it a Visual Positioning System, and it achieves centimeter-level accuracy.
Both companies told Guardian Australia that the ground-level scans were not handed directly to Vantor. Instead, they trained Niantic's foundation models. The distinction matters legally, though it may feel thin to players who submitted scans expecting them to improve AR features, not military drone guidance.
“The fusion of Niantic's ground-level visual data with our aerial mapping software allows us to maintain operational precision even when satellite signals are completely jammed.”
— Vantor spokesperson
What does the Niantic-Vantor partnership actually do?
The technical integration combines Niantic's ground-level Visual Positioning System with Vantor's Raptor software, which incorporates over two decades of satellite imagery. The result is a shared coordinate system for drones, ground vehicles, and AR headsets that works whether GPS is available or not.
Early tests showed error reduction of up to 70 percent and accuracy of about 1.5 meters, according to a Niantic Spatial blog post. The visual system is immune to standard signal jammers because it relies on camera feeds matching against known terrain, not satellite triangulation.
Separately, in February 2026, Vantor landed a US Army contract worth up to $217 million for the One World Terrain program. That effort focuses on high-precision 3D terrain data for the Army's Synthetic Training Environment, covering realistic simulation and mission rehearsal. There's no public evidence that Pokémon Go scans will be part of that specific contract, but the timing underscores Vantor's growing Pentagon ties.
Why are players upset about the data use?
Discussion across Reddit and Hacker News focuses on what many call an "unethical bait-and-switch." Players expressed shock that their casual game contributions were repurposed for military drone guidance. Niantic insists the process uses trained models rather than raw images, but that distinction hasn't satisfied critics who argue they never consented to defense applications.
The legal ground is clear enough. Niantic's terms of service and privacy policies at the time allowed broad use of submitted data, and participation was voluntary. But the ethical question is murkier. When you scan a park bench to unlock a virtual Pikachu, should you expect that scan to eventually help an autonomous drone navigate a combat zone?
Niantic recently split its gaming division from its geospatial AI business. In March 2025, Saudi-backed Scopely acquired the games side for $3.5 billion, while Niantic Spatial continues as a standalone company focused on spatial AI models. That separation may have been strategic, creating distance between the consumer brand and defense applications.
What happens next for crowdsourced spatial AI?
GPS denial is becoming standard on modern battlefields. Russia and Iran use jamming and spoofing routinely. The demand for GPS-free navigation will only grow, and Niantic Spatial has one of the most comprehensive ground-level visual datasets ever assembled, built by an unpaid volunteer army that thought it was playing a game.
Other companies collecting spatial data through consumer apps should expect scrutiny. Any AR application that maps real-world environments, from navigation apps to smart glasses, could face questions about downstream military use. The Pokémon Go case sets a precedent that will be hard to ignore.
Logicity's Take
This is the clearest example yet of how consumer data pipelines can feed military applications without users realizing it. Niantic's terms likely hold up legally, but the reputational damage to AR gaming could be significant. Future crowdsourced mapping efforts will face harder questions from users who now understand where their scans might end up.
Frequently Asked Questions
Did Pokémon Go give player data directly to the US military?
No. The AR scans trained Niantic Spatial's foundation models. Those models power the Visual Positioning System now used in a partnership with defense contractor Vantor. The raw scans were not handed to Vantor or the military.
Was the data collection disclosed to players?
Yes. Participation was opt-in, and Niantic collected permissions under its privacy and terms-of-service policies at the time. However, military applications were not explicitly mentioned.
How accurate is GPS-free drone navigation using this technology?
Early tests showed error reduction of up to 70 percent and accuracy of about 1.5 meters, according to Niantic Spatial.
Why does the military need GPS alternatives?
GPS jamming and spoofing are already used in conflicts in Ukraine and Iran to disable drones and guided missiles. Visual navigation systems are immune to these electronic attacks.
Another case of consumer technology creating unexpected security implications
Need Help Implementing This?
If your organization is evaluating spatial AI, GPS-denied navigation, or the data governance implications of crowdsourced mapping, contact our team for analysis and vendor assessments tailored to your sector.
Source: The Decoder / Maximilian Schreiner
Huma Shazia
Senior AI & Tech Writer
Related Articles
Browse allZuckerberg's Superintelligence Lab Faces Setback
The first AI model from Zuckerberg's superintelligence lab has failed to impress compared to its rivals, sparking concerns about the lab's direction. We take a closer look at what happened and why it matters.

Muse Spark Launch Propels Meta AI App to Top 5
The recent launch of Muse Spark has significantly boosted the popularity of Meta AI app, pushing it into the top 5. We explore what this means for the AI landscape.

Meta's Muse Spark AI Model Lags Behind ChatGPT and Claude
Meta's Muse Spark AI model still can't outperform ChatGPT and Claude in key areas, despite its advancements. We explore what this means for the AI landscape.

Meta Launches Muse Spark AI To Challenge ChatGPT
Meta launches Muse Spark AI to challenge ChatGPT and Claude, we explore what this means for the AI landscape. Muse Spark AI is a significant development in the AI chatbot space.


