Anthropic is exploring a new issue with legal institutions: how AI agents can negotiate to secure the best deals for human users. Two sources familiar with the matter revealed that Anthropic is consulting with law firms to study how lawyers can protect their clients' interests in negotiations while ensuring fair outcomes. A person familiar with the project stated that the company is exploring 'negotiation norms' for model interactions. As intelligent systems evolve, AI agents will represent users in negotiations with other agents that have conflicting interests. The source also mentioned that the company is considering the extent to which agents should rely on human feedback during negotiations, as well as the level of vigilance they should maintain when facing other AI opponents to avoid falling into paranoia and conspiracy theories. Another person close to the company disclosed that Anthropic has also discussed having users sign documents to clarify the entrusted matters and rules that the agents must follow.