Embark says Arc Raiders’ robots don’t learn “on player data” with devs watching “TikToks and YouTube clips” to find new ways to get you

Published:

Last Updated on 16 April 2026

By:

, ,

Prefer FRVR on Google

Reading Time: 2 minutes

There are a lot of misconceptions surrounding Embark Studios’ extremely popular extraction shooter, Arc Raiders. While the game makes use of machine learning to help its robotic ARCs traverse the complex environments of its in-game maps, the robotic enemies aren’t actively learning in real-time.

This is a misconception that has surrounded the game since launch, likely due to certain pop-up surveys that appear in game. However, the actual machine learning technology that powers the movement of the game’s enemies only really applies to the fluid movement animations that help the virtual machines stably walk, jump, and fly around the game.

Martin Singh-Blom, Embark’s machine learning research lead, explained that the robots are not “training on player data”. Even as the developers introduce powerful new creatures to fight, their behaviour won’t be powered by some machine-learned tricks, but instead by the will of developers trying to create something unique.

In a discussion with GamesRadar during GDC, the machine learning expert explained that the tweaks required to make the robots deadlier come from developers watching gameplay and figuring out how to allow its titular ARCs to better fight players.

“We keep trying to make the flying drones better and better at flying indoors,” he explained as one example. “And every time players start to thing that they’re safe in some place, then it’s nice if we can figure out a way to make the drones enter that place”.

The developer explained that one major addition to the movement of ARCs came in the form of the Leaper. On launch, developers saw streamers hiding in a tunnel that the Leaper couldn’t navigate, which caused the team to create more intelligent pathing and detection for the robot to make sure that space was no longer safe.

“It’s not like we’re training on player data,” the developer continued. “The loop is, all the game developers look at TikToks and YouTube clips and say, ‘hey, that’s really cool’. Could we make this change that would really make people squeal? And then we make the change, and then people hopefully get surprised and have a good time.”

As a game with such a unique group of enemies, with even more enemies on the way, Singh-Blom explained that the team is always working on AI improvements. “We hope to have more things in the future and have things in the pipeline,” he said, teasing additional AI improvements to the game, “but they’re not ready yet”.

As far as the Arc Raiders developers are concerned, there’s a back and forth design to the game’s titular enemies. As players find new ways to gain the advantage against the game’s robotic threat, Embark will increase their intelligence, which causes players to find new ways to fight. It’s a cycle that is not only satisfying for players, but also for the developers, as there’s always a new challenge to overcome.

With 30% of Arc Raiders players hyper-focused on PvE play, there are more players working together agaainst the ARCs than expected. With this in mind, Embark may be looking to make the robots even more dangerous as players continue to band together.