Anthropic is consulting law firms to establish fair negotiation norms for AI agents before they begin conducting commercial transactions. The company has confirmed it will haggle over prices, resolve complaints, and improve contract offers through automated systems. This initiative follows internal experiments showing that stronger models secure better deals, a capability that could widen societal inequality if left unregulated.
This development addresses a specific data problem in artificial intelligence. While improving models for math and code relies on plentiful, verifiable data, the humanities lack clear-cut right or wrong answers. Anthropic is already hiring Conceptual Reasoning Fellows, specifically lawyers, to test Claude on judgment-heavy questions in law, philosophy, and economics. Justin Curl, a member of technical staff researching law, institutions, and AI at the company, argues that quality is notoriously hard to define in legal contexts. Without these expert evaluations, the system cannot reliably judge fairness in autonomous negotiations.
- Fellows will spend four months testing Claude on judgment-heavy questions
- Compensation for these roles ranges from $100,000 to $167,000
- Focus areas include law, philosophy, cognitive science, economics, and political science
