This research investigates LLMs' capability to control information strategically in non-cooperative settings through gameplay analysis of "The Chameleon" - a hidden-identity game.

Key findings:

• LLMs demonstrate sophisticated information control abilities in competitive environments
• These systems can strategically conceal information from adversaries while revealing information to cooperators
• LLMs show capability to infer hidden information about other participants
• The research highlights critical implications for AI safety and security protocols in multi-agent systems

Why it matters for Gaming: This research provides valuable insights into how LLM-based agents can function in strategic game scenarios, offering potential for more sophisticated NPCs and game AI that can engage in complex social deduction mechanics.

Do LLMs Strategically Reveal, Conceal, and Infer Information? A Theoretical and Empirical Analysis in The Chameleon Game