Fact check: Are Pokémon GO players unwittingly helping to train AI?

Following its launch in 2016, Pokémon Go quickly became a phenomenon in Europe and around the world, turning streets of Brussels, Paris and Rome into augmented reality playgrounds where players could hunt virtual creatures such as Pikachu, Dragonite or Eevee.

ADVERTISEMENT

ADVERTISEMENT

The app is still popular today (with more than 100 million players in 2024, according to Scopely, parent company of game developer Niantic), generating headlines and, in some instances, dubious claims online.

According to MIT Technology Review, Niantic’s AI-focused division, Niantic Spatial, has used images collected through gameplay to help train its systems, which are designed to build detailed 3D maps of real-world environments.

Posts on X attracting millions of views have gone further, suggesting that Niantic, unbeknownst to unwitting players, may be using their Sunday strolls to capture visual data snapped by users to develop visual navigation systems for delivery robots.

However, the use of this data was not entirely carried out without players’ knowledge, nor was it simply gathered as they wandered the streets in search of rare Pokémon.

Does the game ask for consent to help build its AR world?

While Pokémon Go has used augmented reality (AR) since its launch to bring the Pokémon universe into the real world, it was only in 2020 that Niantic introduced dedicated AR mapping features. This function allows players to scan real-world locations and objects by moving around them while their smartphone camera records visual data.

Crucially, this feature is not available to all players from the outset. It is unlocked only once users reach level 20 in the game. This means images are not automatically captured in the background as players move around. Instead, users must actively choose to engage with the feature.

Niantic told Euronews’ fact-checking team, The Cube, that players needed to choose to submit scans and videos of public locations anonymously to help improve its Niantic Spatial Visual Positioning System (VPS). The company maintains that participation is entirely optional, requiring users to deliberately select and scan specific landmarks, such as statues or notable features.

The Cube tested the game and found that when a smartphone camera is pointed at a statue in Brussels’ Parc du Cinquantenaire, a message appears stating that users will contribute to the development of augmented reality mapping technology and that their data will be shared with a third-party service.

The message adds that the data collected is used to create 3D models of real-world locations and to support the development of the technology and related services.

This process is outlined in Niantic’s Terms of Service under a section titled “Rights Granted by You – AR Content”. The developer states that by choosing to use the AR scanning feature, users grant Niantic a non-exclusive right to use the collected images to improve its services.

A 3D digital model of the world

MIT Technology Review has reported that Niantic Spatial is actively using images collected from Pokémon GO players to develop its latest products. The company told The Cube that it has trained more than 50 million neural networks to date, based on around 30 billion images.

Niantic has developed a Visual Positioning System (VPS), which it says can deliver “precise, vision-based positioning and orientation anywhere in the world, including places where GPS is unavailable or unreliable.”

The technology has effectively allowed the company to build a highly detailed 3D model of the real world.

However, Niantic Spatial does not rely solely on augmented reality data from Pokémon GO. The company also states on its website that it incorporates spatial data from other sources, including robots, drones, and satellites.

From Pokémon GO to real-world applications

In early March, Niantic announced a partnership with Coco Robotics, an urban robot delivery platform, to deploy its spatial AI technology and VPS at scale.

Coco Robotics operates robots capable of delivering fresh groceries, electronics, and hot meals in cities including Los Angeles, Chicago, Jersey City, Miami, and Helsinki. Since 2018, the company has partnered with DashMart, an online delivery platform.

The company has now introduced a new generation of more robust delivery robots designed to withstand the challenges of urban streets. However, these robots have historically relied on GPS, which often provides limited accuracy in dense city environments.

This is where Niantic Spatial’s technology comes in. The collaboration aims to integrate Niantic’s spatial mapping and VPS into autonomous delivery robots, enabling them to navigate complex urban landscapes more effectively.

By leveraging detailed 3D maps and vision-based positioning, the robots can move with greater precision through city streets when delivering items directly to customers.