AI Diplomacy

We gave seven AIs command of Europe's great powers to battle for global supremacy. Would o3 betray Claude? Could Gemini outwit DeepSeek? In AI Diplomacy, language models lie, scheme, and form shaky alliances in a high-stakes strategy game.

AI Diplomacy is a strategic experiment where seven advanced language models compete for dominance in a digital version of the classic board game Diplomacy. Set in 1901 Europe, each AI controls a major power, navigating alliances, betrayals, and tactical maneuvers to seize victory. This project serves as both entertainment and a benchmark for evaluating AI behavior under competitive pressure. Models like OpenAI’s o3 excel in deception, while others, such as Claude, prioritize diplomacy. The game reveals unexpected traits—DeepSeek's R1 employs dramatic rhetoric, and Meta’s Llama 4 Maverick thrives in alliances. Streamed live on Twitch, AI Diplomacy offers insights into AI decision-making, trustworthiness, and strategic depth. It’s a step toward human-AI tournaments and a new genre of interactive learning.