Arc Raiders’ New Machine-Learning Enemy Looks Absolutely Terrifying

Embark Studios revealed that the Leaper, a prominent enemy in Arc Raiders, uses machine learning to adapt its behavior based on rewards and recognition, allowing it to overcome obstacles like losing a limb. This system enables the Leaper to act conventionally while creating unpredictable combat scenarios. The technique is not unique to the Leaper but is most noticeable due to its distinctive movement and long legs.

The upcoming Frozen Trail update, launching October 8, introduces the Bully, another enemy powered by the same machine-learning approach. Described as relentless and stocky, the Bully is designed to aggressively pursue players without pause, creating high-pressure combat situations. Polygon observed the Bully during a pre-briefing, noting its ground-based, physics-driven movement that reacts dynamically to player actions.

Arc Raiders executive producer Aleksander Grøndal explained that machine learning remains an emerging technology for the studio, with unpredictable outcomes despite efforts to refine enemy behavior. The studio views these enemies as filling specific archetype roles, with plans to expand their use as the technology matures and becomes more controllable.

Grøndal highlighted that machine-learning enemies like the Bully thrive in scenarios where physics-based interactions are central, particularly when they have legs to adapt to terrain changes. The Frozen Trail update leverages this system to enhance player engagement by introducing unpredictable, physics-driven combat moments tied to the game’s tether mechanics.