Google DeepMind has taken a minority stake in Fenris Creations, the newly independent studio behind EVE Online, to advance AI research in complex virtual worlds.
The partnership involves testing AI models on an offline version of the space MMO, focusing on long-horizon planning, memory, and continual learning without using live player data.
A New Era for Gaming and AI
Fenris Creations, formerly CCP Games, rebranded after a $120 million deal to buy independence from Pearl Abyss, ensuring continued development of EVE Online and projects like EVE Vanguard.
DeepMind CEO Demis Hassabis, a lifelong gamer, highlighted games' role in AI milestones like AlphaGo and AlphaStar, calling EVE Online a perfect sandbox for multi-agent intelligence.
EVE Online, launched in 2003, stands out for its player-driven economy, massive fleet battles, and emergent politics that mimic real-world societies in space.
Why EVE Online is Ideal for AI Training
Unlike scripted games, EVE's chaos of alliances, betrayals, and scams offers unprecedented data on cooperation, competition, and deception at scale.
This move builds on DeepMind's history of using games for AI benchmarks, potentially unlocking insights applicable to real-world systems like traffic management or supply chains.
For gamers, it promises innovative features, such as smarter NPCs or AI-assisted strategies, revealed at Fanfest 2026.
The independence of Fenris allows bolder experimentation, free from parent company constraints, fostering long-term innovation in persistent worlds.
Amid booming AI investments, this partnership underscores gaming's pivot as a frontier for safe, ethical AI development before broader applications.
Ultimately, it matters to everyday people as smarter AI from such simulations could enhance virtual assistants, autonomous systems, and even economic forecasting tools we rely on daily.
Andrew Lee
