Anthropic seeks lawyers’ advice on rules for AI dealmaking, FT reports

An AI agent negotiating for a user may face another agent pursuing conflicting goals. Anthropic is seeking lawyers’ advice on how those exchanges should work, the Financial Times reported on October 18, 2026. Citing two people familiar with the company’s thinking, the FT described consultations with law firms about how attorneys conduct negotiations.

The lawyers’ role offers a specific model: representing a client while honoring fair outcomes. According to PYMNTS’s account of the FT reporting, Anthropic wants to understand how attorneys manage that balance. The focus is therefore not simply on bargaining skill, but on the practices lawyers use when acting for someone else.

The prospective uses include negotiating prices, resolving complaints and seeking better contract offers on a user’s behalf, according to The Decoder, also citing the FT. One source familiar with Anthropic’s work described the company as considering “negotiation norms” for interactions between models representing users whose goals are at odds.

A separate consideration is how an agent should judge its AI counterpart. The source said Anthropic was balancing how suspicious agents should be without becoming “paranoid” or “conspiratorial.” That concern addresses the agent’s attitude toward the other negotiator, rather than how often it checks back with its own user.

Bargaining ability adds another concern. The Decoder says Anthropic’s experiments found that stronger models negotiated better deals, raising the possibility of wider societal inequality. The experimental result concerns negotiation outcomes; the inequality warning describes a possible consequence, not an observed social effect.

The work remains reported deliberation, rather than an announced set of negotiation rules. Anthropic declined to comment when contacted by the FT.