Players mistakenly thought that enemies were learning in real time.
Despite the active use of machine learning, the robot enemies in ARC Raiders do not learn in real time and do not analyze player behavior directly.
As Martin Singh-Blom, head of machine learning research, explained, the technologies are used exclusively for animation and movement of robots — so that they can naturally walk, jump and fly in complex game locations. At the same time, the behavior models of the enemies themselves are created manually by the developers, and are not formed due to player data.
According to Singh-Blom, the team does not use user data training. Instead, the developers carefully study the gameplay — including videos on YouTube and TikTok — and make changes based on them. For example, if players find safe zones, the team tries to change the behavior of enemies in order to deprive users of this advantage.
One such case was the Jumper robot: after the release, streamers discovered a tunnel that it could not reach. As a result, the developers improved the navigation and detection system to eliminate this "loophole".
![[Video] IGN Gave It a 6. We Recommend It - Review of Mouse: P.I. for Hire [Private Detective MAUS]](https://media.ixbt.site/312x192/smart/ixbt-data/1175498/a-first-look-at-mouse-p-i-for-hirerubber-hose-and-rubber-bullets-2025-06-10-394764.jpg)
