A group of major financial institutions, including Bank of America, Capital One, and ING, published a paper titled Building Trust in Agentic Commerce. The document outlines proposed safety principles for AI agents that facilitate payments between consumers and merchants. The banks argue that providers must maintain audit trails of consumer instructions, authentication, and transaction outcomes to resolve disputes and combat fraud.
The report highlights risks associated with AI autonomy, such as agents making purchases without final customer review or falling for scams. Banks note that current payment chains often lack real-time visibility into an agent's identity or the customer's original intent. The proposed framework suggests that customers should be able to manage the authority delegated to these agents through secure, auditable methods.
The paper addresses concerns regarding transparency, noting that agents should disclose how they prioritize products, including sponsored options that might favor higher commissions. It also warns that while detailed logs of shopping prompts and decision-making are necessary for fraud investigations, they create risks related to data privacy and potential profiling if not managed with limited access.
The principles also emphasize interoperability, suggesting that customers and merchants should have the freedom to choose agents and payment services without unreasonable restrictions. The current proposal is voluntary and nonbinding, with no specific implementation timeline. The participating banks plan to release a follow-up paper to discuss translating these principles into industry standards.
