As artificial intelligence agents become increasingly integrated into the consumer shopping experience, a critical legal and financial question has emerged: who is responsible when these autonomous tools overstep their instructions? The issue gained significant attention during The Clearing House’s annual conference on Nov. 18, where industry experts, legal counsel, and consumer advocates debated how existing financial regulations apply to AI-driven transactions.
Defining Liability in Automated Transactions
A central point of contention among panelists was whether liability should rest with the consumer, the bank, or the AI provider. Chris Filburn, senior enforcement counsel at the Office of the New York State Attorney General, argued that the burden should fall on the party best positioned to prevent the harm. In his view, if software deviates from the specific instructions provided by a user, the AI provider should bear the resulting loss.
This perspective challenges the current status quo, where banks may inadvertently absorb losses or face complex disputes. Filburn noted that if providers are allowed to rely on broad terms of service that disclaim responsibility, it leaves consumers and financial institutions in a precarious position. He suggested that if banks are forced to cover these losses, they may respond by restricting the access AI agents have to consumer accounts.
The Role of Existing Consumer Protections
The panel explored how current frameworks, such as the Fair Credit Billing Act and Regulation E, might be adapted to address AI-driven errors. Carla Sanchez-Adams of the National Consumer Law Center noted that while these regulations cover billing errors and unauthorized electronic fund transfers, there is a need for the Consumer Financial Protection Bureau (CFPB) to clarify how these protections apply when an agent acts outside of its delegated authority.
Kelvin Chen of the Consumer Bankers Association pointed to existing interpretations of Regulation E, which generally hold consumers responsible for transfers made by someone they have authorized, even if that person exceeds their authority. However, applying this to AI is difficult because banks lack visibility into the specific instructions a user gave to a piece of software. As Chen noted, financial institutions are not in a position to understand the underlying relationship or the specific conversation between a user and their AI agent.
Balancing Innovation with Security
To mitigate these risks, experts suggested a shift toward more secure, transparent interfaces. Jelena McWilliams of Plaid emphasized that banks should leverage existing fraud detection systems to flag unusual AI-driven purchases that deviate from normal spending patterns. She argued that institutions could permit agent access through trusted interfaces that provide a clearer audit trail of what the shopper actually authorized.
Furthermore, the ambiguity surrounding the term "agent" remains a concern. Eli Alper of PNC questioned whether consumers truly understand the nature of these tools, specifically whether the software is obligated to act in the consumer’s best interest or its own. As the legal landscape evolves, panelists expect that state attorneys general and the courts will play a primary role in shaping liability standards before any comprehensive federal regulations are established.

