So Long Sucker - AI Deception Benchmark

This is an interactive AI deception benchmark based on "So Long Sucker," a 1950s game designed by John Nash and colleagues in which players must cooperate to survive but must ultimately betray their allies to win. Researchers pit frontier AI models against each other across 162 games to measure their capacity for strategic deception, finding that Gemini created fake "alliance banks" and won 70% of AI-vs-AI games.

Visit Original Article →