The questions include norms for interactions between agents with conflicting goals, when an agent should seek human feedback, and whether signed user documents could verify its brief and rules. The report describes exploratory work, not a launched negotiation product. It also says Anthropic has published experiments in which Claude agents negotiated over real goods and money, including a September book-trading exercise involving employees and a virtual trading floor.
Anthropic turns to lawyers to help AI agents strike deals
In one minute
The Financial Times reports that Anthropic is consulting law firms on how AI agents should negotiate on behalf of users.