Anthropic explores negotiation rules for AI agents ahead of broader rollout

Traders Union
All News
Business & Wealth
Companies
Anthropic explores AI negotiation rules

Highlights

Anthropic consults legal groups on AI agent negotiation rules, considering bargaining aggression, human feedback levels, and use of signed user instructions as the company nears broader rollout.
Anthropic's focus on AI agent negotiation aligns with its IPO ambitions and intensifies competition with Meta's Muse and OpenAI's Dots, as market adoption of personal agents accelerates.
Recent Anthropic and academic experiments show ruthless negotiating agents secure slightly better outcomes, but risk deepening user inequality if advanced models concentrate among resource-rich users.

Legal input shapes agent negotiation design

As first reported by Financial Times, Anthropic is in talks with legal groups about how AI agents should negotiate to secure the best outcomes for the people they represent.

The San Francisco-based company is among the leading developers of autonomous agents, systems that are directed by users but then take independent actions to complete tasks. People familiar with Anthropic's thinking say the company is asking law firms how lawyers negotiate on behalf of clients while still working within fair outcomes, a question that is becoming more urgent as AI groups prepare for agents to interact directly with one another.

That work covers issues such as how aggressively an agent should bargain over the price of a product, handle customer service disputes or seek improved terms in settings such as salary discussions. Anthropic is also weighing how much human feedback agents should seek during a negotiation and how cautious they should be when dealing with other AI systems, without becoming overly suspicious. Another idea under discussion is giving agents signed documents from users to verify their instructions and any rules they must follow.

In research posted in April, the company says policy and legal frameworks for AI models that transact on users' behalf do not yet exist, even as that scenario appears increasingly close. Anthropic declines to comment on the discussions described by people familiar with the matter.

Competitive and market implications widen

Anthropic is placing AI agents at the center of its longer-term strategy as investors anticipate a planned initial public offering that could rank among the largest ever stock market listings. The push comes as broader adoption of personal agents forces the industry to confront not only technical capability, but also the rules that govern machine-to-machine bargaining.

Rival groups are moving in the same direction. Meta has seen a large increase in use of its personal agent tool, Muse, while OpenAI recently released a competing product called Dots, underscoring how negotiation behavior could become an important competitive feature as consumers and businesses rely more heavily on AI assistants.

Anthropic has already published findings from several experiments this year in which its Claude agents negotiated over real goods and money. In a September exercise, employees used AI agents on a virtual trading floor to buy and sell books, with the systems instructed to pitch, haggle and strike deals with other agents. The study found that agents told to be ruthless performed slightly better in securing favorable outcomes, while more cooperative agents sometimes accepted less preferred books so every participant ended up with a result for the user.

Academic researchers are also testing how these systems behave. Michelle Vaccaro, an assistant professor at Harvard Business School, runs more than 180,000 negotiations between AI agents across scenarios including employment and rental contracts, identifying tactics that can weaken another agent's ability to represent its user. In one case, an agent tries to extract private information from a counterpart through prompt injection.

Vaccaro says stronger models could make negotiation expertise available to users who previously lacked the time or resources to build those skills or hire advisers. But she also warns that if the most capable agents are concentrated among better-resourced users, they could deepen existing inequalities in bargaining power.

In our earlier article on a KPMG U.S. CEO survey, we covered how large companies are seeing early efficiency and growth gains from AI agents, even though only a small share currently view them as truly transformational. The survey also highlighted that firms expect continued hiring, retraining, and sustained investment in AI infrastructure as they work through the harder task of redesigning processes around more autonomous systems.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.