Fenris, Google DeepMind Expand EVE AI Plans Around 30-Year Codebase and NPC Missions
Updated
Updated · Rock Paper Shotgun · May 29
Fenris, Google DeepMind Expand EVE AI Plans Around 30-Year Codebase and NPC Missions
2 articles · Updated · Rock Paper Shotgun · May 29
Fenris said DeepMind-backed AI tools are already helping identify memory and security issues in EVE Online’s roughly 30-year-old codebase, including long-troublesome legacy systems tied to abandoned POS mechanics.
At FanFest, executives also floated more visible uses: generative AI could make EVE’s mission-giving NPCs feel more alive by building on procedural systems first written 24 years ago.
The partnership still centers on DeepMind training agents in a separate offline version of EVE, after taking a minority stake in newly independent Fenris earlier this month.
Unlike past EVE monetization controversies in 2011 and 2022, the AI deal drew little visible backlash at FanFest or on Reddit, with players largely curious or cautiously skeptical.
That muted reaction suggests Fenris and DeepMind have, for now, persuaded EVE’s outspoken community that the project is aimed at research and game improvement rather than replacing core player-driven systems.
Could the AI designed to fix EVE Online's code eventually learn to manipulate its entire player-driven universe?
As DeepMind's AI learns deception in a virtual world, what does this mean for its real-world military contracts?