Dissociative Identity: Language Model Agents Lack Grounding for Reputation Mechanisms

arXiv·medium signal

Argues that extending human identity verification (KYC, credit scores) to AI agents is fundamentally incomplete. Reputation mechanisms assume continuity of identity, but LLM agents can be trivially cloned, forked, or reset, undermining the basis for trust in the emerging agentic web. Important framing for agent-to-agent commerce and trust protocols.