Anthropic Consults Law Firms on Fair Negotiation Norms for AI Agents

A company called Anthropic is working with law firms to establish guidelines for how artificial intelligence agents should negotiate on behalf of users. This effort aims to ensure that these negotiations are fair and equitable, particularly in areas where clear-cut answers may not exist. According to reports, the company has already demonstrated through experiments that more advanced AI models can secure better deals, which could potentially exacerbate existing social inequalities.

Anthropic’s push for establishing negotiation norms is driven by a recognition of the challenges involved in developing AI agents capable of handling complex tasks like law and philosophy. Unlike math and code, where verifiable data is readily available or easily generated, these areas often lack clear-cut right or wrong answers. As a result, Anthropic needs experts who can create test cases and evaluations to strengthen its Claude model.

The company has already begun addressing this issue by hiring Conceptual Reasoning Fellows – lawyers with expertise in law, philosophy, cognitive science, economics, and political science. These fellows will spend four months testing Claude on judgment-heavy questions for a salary of $100,000 to $167,000 over the course of their tenure.

Justin Curl, an Anthropic researcher focused on AI’s intersection with law and institutions, notes that defining quality in these areas is notoriously difficult. This challenge underscores the complexity of developing AI agents capable of navigating nuanced domains like law.