Intentional Deception as Controllable Capability in LLM Agents

Jason Starace, Terence Soule

As LLM-based agents increasingly operate in multi-agent systems, understanding adversarial manipulation becomes critical for defensive design. We present a systematic study of inte…