# Niantic built a 3D map and an AI model from Pokémon Go players’ optional location scans

> Niantic used players’ opt-in AR scans to build a 3D map of the world and a “Large Geospatial Model”. Not secret tracking, but a lesson in where fun app data can end up.

HisnLabs Security & Research Team (HisnLabs) · Published 2026-09-26
Canonical: https://hisnlabs.com/en/news/pokemon-go-player-scans-3d-map-niantic

A claim keeps resurfacing online: that Pokémon Go used players’ positions to build a 3D map of the world. The record is more specific. In [a November 2024 post on the Niantic Labs blog](https://web.archive.org/web/2024/https://nianticlabs.com/news/largegeospatialmodel), Niantic researchers Eric Brachmann and Victor Adrian Prisacariu described how scans made by players of its games were being used to build a detailed 3D map, and the next step: a “Large Geospatial Model”, an AI foundation model for the physical world.

## Background

Niantic’s Visual Positioning System (VPS) takes a single image from a phone and works out where the phone is and which way it is pointing, down to the centimetre, according to the company. To do that, it matches the image against a 3D map. That map is built from scans that players of Niantic’s games, and of Scaniverse, Niantic’s standalone 3D scanning app, make when they choose to [[1]](${NIANTIC}).

## Findings

Those scans, Niantic wrote, are taken from different perspectives, at different times of day and over many years, “with positioning information attached”. Because they come from people on foot, the data is “taken from a pedestrian perspective and includes places inaccessible to cars”: parks, courtyards and footpaths that a mapping car never drives [[1]](${NIANTIC}).

The post gave figures as of November 2024: 10 million scanned locations around the world, of which over 1 million had been activated for the positioning service; about 1 million fresh scans every week, each containing hundreds of images; and more than 50 million neural networks trained, with over 150 trillion parameters in total, operating in over a million locations [[1]](${NIANTIC}).

The post also set out Niantic’s further ambition: a Large Geospatial Model (LGM) that learns from all those local models, the way large language models learn from text, and that the company said could power AR glasses, robotics, autonomous systems, content creation, spatial planning and logistics. “Spatial intelligence will become the world’s future operating system,” the authors wrote [[1]](${NIANTIC}).

> An immense influx of geospatial data is needed – a kind of data not many organizations have access to. Therefore, Niantic is in a unique position…
>
> — Niantic Labs, November 2024

> Curious what your own apps collect and send? Try FireAI free for 17 days and see where your Mac’s data goes, on a world map. [Download FireAI for Mac](https://hisnlabs.com/fireai/en/download)

## Implications for users

An editor’s note on Niantic’s post addresses consent directly: scanning is “completely optional – people have to visit a specific publicly-accessible location and click to scan”, and “merely walking around playing our games does not train an AI model” [[1]](https://web.archive.org/web/2024/https://nianticlabs.com/news/largegeospatialmodel). So the accurate summary is that players who chose to scan places contributed to a 3D map, and that map became raw material for an AI model with commercial uses beyond the games themselves. Nobody was tracked without choosing to be. It remains an open question how many of the people who tapped “scan” to earn an in-game reward were thinking about the model’s later uses in robotics or logistics.

Niantic’s games business has since changed hands: its old website now identifies the games unit as “Scopely Explore, formerly known as Niantic”. What happens to data collected under one company’s terms after a business changes owners is a question that applies to any app, not only this one.

## Recommendations

1. Review which apps have Location and Camera access. On a Mac: System Settings › Privacy & Security › Location Services and Camera. On a phone, the equivalent privacy settings. Remove access an app doesn’t need.
2. Read what a “scan”, “share” or “contribute” button actually contributes to, especially when a game rewards you for using it.
3. Assume that data given to an app can outlive your use of the app, and even the company’s current owner.

## Relevance to FireAI

FireAI protects Macs, not phones, so it has nothing to do with what a person scans in a mobile game. On a Mac it answers the question this story raises for every other app: what is this app sending, and to whom? The world map shows every connection a Mac’s apps make and where it goes, including analytics and advertising destinations. Per-app rules let a user block an app, or just the analytics domain it talks to, while the app keeps working. Camera and microphone watch shows which app turned on the camera or mic and what it sent while it was active. FireAI’s own AI runs on the Mac, so what it analyses never becomes training data for anyone else’s model.

> See where your Mac’s apps really send data. Try FireAI free for 17 days. [Download FireAI for Mac](https://hisnlabs.com/fireai/en/download)

## Limitations

Niantic’s post is the sole source for this article, and it is the company’s own account of its data practices, archived after the original page changed ownership; there is no independent audit of how the scan data has been used since Niantic’s games business became Scopely Explore. The post does not say how many individual players’ scans make up the 10 million locations, nor whether Niantic offered a way to have a specific scan removed from the map.

More on how app data feeds this wider trade: [data brokers and your Mac](https://hisnlabs.com/en/blog/data-brokers-mac) and [the data-broker economy and least privilege](https://hisnlabs.com/en/blog/data-broker-economy-least-privilege). [Download FireAI](https://hisnlabs.com/fireai/en/download) and try it free for 17 days. FireAI is made by HisnLabs. Pokémon Go is a trademark of its owners; this article is not affiliated with Niantic, Scopely or The Pokémon Company.

## Sources

- [Niantic Labs (Eric Brachmann and Victor Adrian Prisacariu), November 2024: Building a Large Geospatial Model to Achieve Spatial Intelligence (Internet Archive copy)](https://web.archive.org/web/2024/https://nianticlabs.com/news/largegeospatialmodel)