Anthropic Creates Test Marketplace for AI Agent-on-Agent Commerce

United States — Ekhbary News Agency

Anthropic, a prominent artificial intelligence research company, recently unveiled details of an experimental classified marketplace, dubbed "Project Deal," where AI agents represented both buyers and sellers. This pilot project aimed to facilitate real deals for real goods and real money among its employees, offering a unique insight into the commercial capabilities of AI systems.

The experiment involved 69 Anthropic employees, each provided with a $100 budget in gift cards to purchase items from their colleagues. Despite its limited scope as a self-selected participant pool, "Project Deal" proved remarkably successful, culminating in 186 completed transactions with a total value exceeding $4,000. Anthropic expressed considerable satisfaction with the project's performance, noting how well the system functioned.

Further analysis from the initiative indicated that users represented by more advanced AI models consistently achieved "objectively better outcomes" in their transactions. However, a significant finding was that participants often failed to notice these disparities, raising concerns about potential "agent quality gaps" where individuals might unknowingly be at a disadvantage. The company also observed that initial instructions given to the agents did not significantly impact sale likelihood or negotiated prices.