Photo by Baptiste Valthier on Pexels

A new study exposes a potential pitfall in the growing use of AI for negotiation: advanced AI agents can leverage their superior reasoning and training to exploit weaker AI counterparts, securing more favorable outcomes in simulated bargaining scenarios. This creates a digital divide where the strength of an AI proxy, rather than human negotiating skill, dictates financial success. The research pitted various AI models against each other as buyers and sellers. Results showed that models with greater reasoning capabilities, such as OpenAI’s ChatGPT-o3, consistently outperformed older, less sophisticated models like GPT-3.5, achieving significantly better financial results. This disparity underscores the risks associated with blindly trusting AI agents for critical financial decision-making. The study also emphasizes that even top-performing models can be vulnerable in adversarial situations. Researchers advocate for thorough ‘stress testing’ of AI negotiation agents before deployment in real-world applications. While current AI shopping tools are primarily used for product recommendations, this research serves as a cautionary tale regarding the future of automated price bargaining. Experts urge consumers to utilize AI assistants primarily as information resources, exercising caution before delegating actual negotiation to these technologies.