Banks are testing AI agents that can make purchases, move money and answer customers. As the technology spreads, a basic question remains unsettled: when an agent makes a mistake, who pays? Industry groups have floated answers, but no U.S. rule yet settles it.
Key Highlights
- Banks, payment companies and technology firms are working on liability standards for AI-agent transactions, according to Digital Today.
- Former regulators say banks using AI chatbots can still face liability under existing consumer-protection laws, American Banker reported.
- Federal bank regulators excluded agentic AI from revised model risk guidance in April, American Banker said.
- The proposals described here are industry ideas, not law.
The Core Dispute
The central issue is who is responsible when an AI agent errs or is defrauded: the consumer, the bank or the AI developer, Digital Today reported on October 14.
Lead Bank has proposed a “waterfall” framework that assigns liability by cause. Under it, consumers would not be liable if an AI exceeded the authority it was given, while banks would be responsible if they failed to control that authority. Tech firm Primitive suggested a protocol to verify an AI’s identity, the user’s approval and a transaction’s traceability.
American Banker reported that the framework also suggests where banking regulations may need to change. Getting rules to align with agent-driven activity, one executive said, is the bigger challenge.
What Existing Law Says
Former regulators argue the law already applies. In a report covered by American Banker on October 15, Adam Teitelbaum, a former consumer-protection director in the D.C. attorney general’s office, and Stephanie Nguyen, a former Federal Trade Commission chief technologist, said banks may break consumer-protection laws if chatbots give false information or block customers from help. Teitelbaum said corporations “cannot hide behind AI to dodge accountability.”
The report says state attorneys general and individual consumers can act under state laws. American Banker noted that the Conference of State Bank Supervisors released an AI supervisory framework last month for examiners.
A law firm analysis offers a similar view. Baker McKenzie wrote in June that courts and regulators will generally look to the humans and companies behind an AI agent as responsible for its actions.
Gaps in Federal Oversight
American Banker reported that the Office of the Comptroller of the Currency, the Federal Reserve and the FDIC expressly excluded agentic AI from interagency model risk guidance revised in April. Experts told the publication that regulators still expect strong compliance and risk programs. The Consumer Financial Protection Bureau has largely halted enforcement under the Trump administration, the outlet added.
Payments rules add uncertainty. Banking software firm Backbase wrote that Regulation E caps consumer liability for unauthorized transfers at $50 if reported within two days. It said regulators have not updated the rule for agent-initiated payments, and it remains open whether an AI agent counts as an authorized delegate. That is a vendor’s reading, not a regulator’s.
Who Owns the Risk?
Some executives say the answer is plain. Fredrik Davéus, CEO of Kidbrooke, told FinTech Global that an institution that builds AI into its advice owns how that AI is used.
Adoption is uneven. American Banker reported that small and regional banks have been cautious about using agentic AI at scale because it needs ongoing oversight, while large banks use hybrid systems with less human involvement.
What Remains Uncertain
No federal rule defines liability for agent-initiated transactions. The Lead Bank and Primitive ideas are proposals that regulators have not adopted. How courts would treat an agent acting beyond its instructions is also untested, based on the reporting reviewed.
For another angle on how AI is rattling financial markets, see our report on the Wall Street AI sell-off.
What to Watch Next
- Whether federal regulators issue guidance on agentic AI.
- State enforcement and the CSBS framework in practice.
- Whether banks adopt liability frameworks such as Lead Bank’s.
Sources: Digital Today, American Banker, FinTech Global, Baker McKenzie, Backbase. This article will be updated as more information is confirmed.