← Back to News
Innovation

Fed’s Waller calls trust the biggest barrier to agentic commerce

Federal Reserve Governor Christopher Waller told the Sibos conference in Miami that trust is the biggest barrier to agentic commerce, naming authentication, liability, and fraud as the challenges to solve. He spoke for himself and announced no rules or standards.

Networks mentioned

Federal Reserve Governor Christopher Waller said on September 29, 2026, that the biggest barrier to scaling agentic commerce, the use of AI agents to make e-commerce purchases, is “building sufficient trust among buyers and sellers.” Speaking at the Sibos conference in Miami, he named three concrete challenges: authentication, liability, and fraud. The agent-delegated model, where the agent shops and pays on the buyer’s behalf, requires “a more extensive buildout of trust mechanisms and guardrails,” he said.

The speech, “Payments in the Age of AI Agents,” contains no announcement: no rule, no guidance, no standard. A footnote states that the views are Waller’s own “and not necessarily those of my colleagues on the Federal Reserve Board or the Federal Open Market Committee.” Waller “stopped short of prescribing a specific standard,” PYMNTS wrote on September 30.

3
trust challenges named: authentication, liability, and fraud
Federal Reserve Board
92%
fewer sanctions screening false positives with LLMs
FEDS 2025-092
11%
increase in detection rates, in the same study
FEDS 2025-092
19
industry representatives at the Fed’s May 26, 2026, roundtable
Federal Reserve Board

Proving the agent’s authority, not only the payer’s

Authentication comes first. “The question shifts from proving that a buyer is an authorized payer to proving that an agent has the authority to pay on the buyer’s behalf,” Waller said, adding that capturing this “will require new authentication approaches.” Liability, in his words, “comes down to a pretty simple question: Who is on the hook if an agent makes the wrong purchase?”

Existing e-commerce liability frameworks, including network rules and consumer protection standards, could potentially be adapted, he said, but there may also be room to test new approaches. One example is technical standards designed to give everyone a better understanding of what the buyer intended and how the agent carried it out, which could help reduce the ambiguities behind many of today’s transaction disputes. On fraud, Waller warned that detection and prevention systems, “calibrated to human behavior, may not translate well to agents,” so fraud models and rules will need recalibrating for agent payment patterns.

🔑
Assisted versus delegated agents
In the agent-assisted model, the buyer uses an AI agent for product search but makes the decisions and handles the payment. In the agent-delegated model, the buyer grants the agent authority to shop and pay, possibly within constraints the buyer sets, for instance with a pre-funded virtual card to buy groceries online based on past purchases. Consumer-to-business transactions are seen as the first wave, with assisted uses preceding delegated ones, which carry higher risks of unintended purchases.

B2B purchasing suits agents but raises the stakes

Business-to-business purchases “may be especially well suited for agentic commerce,” Waller said. They are often recurring and follow rules such as approved suppliers and budget limits, which give agents natural guidelines. Agents could also negotiate terms with suppliers and optimize working capital. Higher B2B transaction values amplify the exposure to agent errors or unauthorized actions, however. B2B agents would need to pay over ACH, wire, instant payments, and cards, while consumer e-commerce relies mostly on cards.

The glass facade and canopy of the Miami Beach Convention Center, lined with palm trees, on a sunny day
Sibos 2026 runs from September 28 to October 1 at the Miami Beach Convention Center, under the theme “Digital finance for AI-driven economies.”

Platform-specific or interoperable standards

Tech firms, e-commerce platforms, and payment service providers are building protocols that standardize how agents interact with merchants and pay, Waller said, without naming any company. Card networks are publishing specifications to register agents and log cardholders’ approvals. PYMNTS pointed to Visa and its Intelligent Commerce technology, and to Mastercard and its Agent Pay infrastructure.

“A key question is whether agentic commerce will migrate toward platform-specific or interoperable standards,” he said. Interoperable standards, designed to work across e-commerce systems, agent interfaces, and payment methods, could help level the playing field for smaller merchants and payment providers. A related choice sets open systems, which accept a range of shopping agents, against closed ones, where a retailer might require its own. “At this early stage, it is not clear which model will have the edge,” Waller said.

LLMs to cut false positives in sanctions screening

Opening on cross-border payments, Waller said the contextual awareness of large language models “can greatly improve the accuracy of sanctions screening and anti-money-laundering systems.” In a footnote, he cited a September 2025 Fed staff working paper by Jeffrey Allen and Max Hatfield. On average, across realistic matching thresholds, LLMs cut false positives by 92% and raised detection rates by 11% against the best fuzzy-matching baseline, but ran more than four orders of magnitude slower. Given the velocity of modern payment systems, Waller said, “I expect LLMs to augment, rather than replace, faster, more traditional anomaly detection methods.”

From a May roundtable to open questions

The speech follows a roundtable on agentic commerce that Waller hosted at the Fed on May 26, 2026, with Governors Michael Barr and Lisa Cook and 19 industry representatives, including from Amazon, Google, Stripe, and Visa, according to the Board’s meeting summary. Market participants “broadly agree that agentic commerce is in an early phase,” Waller told the Miami audience. He closed with open questions, including which standards are still missing for agents to carry identity, consent, and payment credentials, and how much “purposeful friction” agent authorization should keep.

September 2025
Fed staff study on sanctions screening
A working paper finds LLMs cut false positives by 92% against fuzzy matching.
September 29, 2025
Sibos in Frankfurt
Waller covers distributed ledgers, stablecoins, and AI.
May 26, 2026
Roundtable at the Fed
Governors Waller, Barr, and Cook meet the industry on agentic commerce.
September 29, 2026
Sibos in Miami
Waller names three challenges: authentication, liability, and fraud.

Provenance

Published September 29, 2026

7 sources, 4 distinct domains

↗ Federal Reserve Board, Christopher J. Waller, “Payments in the Age of AI Agents,” Sibos 2026, Miami · federalreserve.gov↗ PYMNTS, Fed’s Waller Says AI Shopping Changes What Banks Need to Authenticate · pymnts.com↗ PaymentsJournal, Federal Reserve at Sibos: Gov. Waller on Agentic AI · paymentsjournal.com↗ Federal Reserve Board, Payments Innovation Roundtable meeting summary, May 26, 2026 · federalreserve.gov↗ Federal Reserve Board, Can LLMs Improve Sanctions Screening in the Financial System? (FEDS 2025-092) · federalreserve.gov↗ Federal Reserve Board, Christopher J. Waller, “The Next Frontier of Payments Innovation,” Sibos 2025 · federalreserve.gov↗ Sibos 2026, Miami Beach Convention Center, September 28 to October 1 · sibos.com
← All news