Pokémon Go data trained AI that could assist military drones in war zones

· Source: AI (artificial intelligence) | The Guardian · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Cybersecurity & Data Privacy · Depth: Fundamental Awareness, short

Summary

AI models designed to assist military drones in war zones with navigation have been trained using location scan data collected from users of the popular augmented reality game Pokémon Go. Launched in 2016, Pokémon Go introduced a 2021 update allowing players to voluntarily upload real-world scans for in-game rewards, accumulating over 800 million downloads by 2018. Niantic, the game's creator, used these historical scans to train its AI to interpret physical spaces. Niantic Spatial, a spin-off, partnered with Vantor in December to apply this technology for drones, enabling precise navigation in GPS-unavailable or compromised environments. While companies state raw game scans weren't directly provided to Vantor, experts like Tom Sulston and Dr. Rob Nicholls raise concerns about civilian data exploitation for military purposes, citing similar cases like Strava data. Vantor also secured a US Army contract worth up to US\$217m in February, and Niantic sold its gaming division to Scopely for US\$3.5bn in 2025.

Key takeaway

For policy makers and AI ethicists evaluating data privacy and the repurposing of user-generated content, this case highlights the urgent need for "best interests of the user" or "fair and reasonable" tests in data governance. Your current regulatory frameworks may not adequately protect civilian data from being used for military or other unintended applications, even with opt-in clauses. Consider strengthening regulations to ensure transparency and explicit consent for all potential data uses, especially when "free" services treat users as the product.

Key insights

Pokémon Go user location data trained AI now leveraged for military drone navigation in GPS-denied environments.

Principles

Method

Niantic collected user-uploaded real-world scans via Pokémon Go's Pokéstop feature, then used this data to train foundation AI models for spatial recognition.

In practice

Topics

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Ethicist, Policy Maker, General Interest

Related on AIssential

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI (artificial intelligence) | The Guardian.