Reporting citing the Financial Times says Anthropic is discussing fair-negotiation norms with law firms for AI agents handling price bargaining, complaints, and improved contract terms. Unlike code or mathematics, law and social science lack easily verifiable answers, creating a need for experts to develop evaluation cases and quality criteria. The report also notes that access to stronger models could give some users a negotiating advantage. Applicable rules and real contract outcomes have not yet been disclosed. The discussions should be understood as early efforts to define an agent’s authority and permissible negotiating behavior.