Banks Call for Audit Trail From AI Shopping Instructions to Payment Outcome

Six major banks have published a proposal urging artificial intelligence shopping transactions to maintain a detailed record of customer instructions, authentication and the transaction's outcome. This proposal, titled "Building Trust in Agentic Commerce," was published on September 22 and was written by ASB Bank, Bank of America, Capital One, Commonwealth Bank of Australia, ING and NatWest. The banks argue that these records are vital for investigating scams, recovering lost money, and resolving disputes.

The term "agentic commerce" refers to the use of AI agents in facilitating payments between consumers and merchants. While an AI agent may initially search for a product, it could potentially select and purchase the item without the customer reviewing the final choice. As agents gain more autonomy, the risks associated with these transactions increase.

The proposed safety principle would provide customers with the ability to view and manage the authority granted to agents. This includes secure and auditable methods of entering payment credentials and authorizing purchases. Any party facing potential liability could require authentication and establish dispute processes to involve the relevant participants. Liability would reflect where an error or risk originated within the transaction.

Safety concerns arise from agents asking for card details, entering them directly into websites or favoring payment methods with weaker protections. Criminals could potentially impersonate or compromise agents and merchants. Information sharing, customer warnings, fraud interventions and recovery efforts throughout the payment chain are necessary to address these issues.

The banks emphasize the need for transparency, such as explaining how agents prioritize products and payment methods. They also call for limited access to data, ensuring that each participant only has the necessary information for their respective functions, with consent for any additional uses or sharing. The remaining principles in the proposal address choice and interoperability, allowing customers and merchants to select agents, wallets and payment services without unreasonable restrictions while still permitting providers to choose which services they support based on safety, commercial and legal grounds.

This proposal, though voluntary and nonbinding, aims to ensure trustworthy agentic commerce by reconstructing the path from a customer's instruction to the resulting transaction and using the record when issues arise.