Anthropic tests Claude in EVE Online

Anthropic researchers tested Claude AI models in massively multiplayer online role-playing games (MMORPGs) like EVE Online and EVE Frontier to study how language models handle profit-seeking markets and multi-agent coordination. Researchers found that newer models successfully executed low-risk market spread-capture strategies and coordinated simple team tasks, but struggled with directional trading and often failed to detect hidden saboteurs within their teams. Open article →