Anthropic is consulting with legal institutions to solve a new problem: how AI agents should negotiate to obtain optimal transaction results for human users. Two people familiar with Anthropic's thinking say the company is seeking advice from law firms to study how lawyers can conduct negotiations that both represent clients' interests and balance fair results. A person familiar with the Anthropic project said the company is exploring “negotiation specifications” when models interact. As intelligent systems continue to be iterated, AI agents will represent users and connect with other agents with conflicting targets. The source added that the company is still weighing: the extent to which intelligence agencies need to seek human feedback during negotiations; how vigilant they should be in the face of other AI opponents while avoiding falling into questionable and conspiracy judgments. Another person close to the company revealed that Anthropic has also explored a solution: let the user issue a signed document to confirm the entrustment matters and all the rules that need to be followed with the agent.

Zhitongcaijing · 2d ago
Anthropic is consulting with legal institutions to solve a new problem: how AI agents should negotiate to obtain optimal transaction results for human users. Two people familiar with Anthropic's thinking say the company is seeking advice from law firms to study how lawyers can conduct negotiations that both represent clients' interests and balance fair results. A person familiar with the Anthropic project said the company is exploring “negotiation specifications” when models interact. As intelligent systems continue to be iterated, AI agents will represent users and connect with other agents with conflicting targets. The source added that the company is still weighing: the extent to which intelligence agencies need to seek human feedback during negotiations; how vigilant they should be in the face of other AI opponents while avoiding falling into questionable and conspiracy judgments. Another person close to the company revealed that Anthropic has also explored a solution: let the user issue a signed document to confirm the entrustment matters and all the rules that need to be followed with the agent.