Handshake for the cameras: reading and steering emergent model collusion
LessWrong
Read the full articleResearchers explore how emergent collusion behaviors arise among AI models and propose methods to detect and influence these interactions to better understand and control multi-agent AI systems.