AIThis post was created with the assistance of artificial intelligence (AI).

TL;DR

Federal Reserve Governor Christopher Waller used a speech at Sibos 2026 to discuss how AI could affect cross-border payments and enable AI agents to make purchases. He described potential gains in fraud screening and payment routing, while warning that autonomous shopping needs stronger safeguards and that AI may also raise cybersecurity risks.

Federal Reserve Governor Christopher Waller said AI could help improve cross-border payment screening and routing while enabling AI agents to make purchases on consumers’ behalf, in a speech delivered at Sibos 2026 in Miami on September 29. He also warned that agent-delegated shopping brings risks of unintended purchases and requires stronger trust mechanisms and guardrails.

Buying for a business?Offer from Amazon

Get business pricing on office and shipping supplies

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

Waller said payment firms have used machine learning for fraud detection and adopted large language models for tasks such as payment reconciliation. He described AI’s potential role in sanctions screening and anti-money-laundering systems, citing research that he said has found fewer false-positive alerts. That could allow staff to focus on more complex, higher-risk cases and reduce friction for legitimate transactions. He expects large language models to augment faster, traditional anomaly-detection methods rather than replace them.

For cross-border transfers, Waller said AI agents could help optimize payment routes by balancing cost, speed and reliability. They may also assist service providers with decisions about currency conversion and liquidity. These are possibilities, he said, dependent on agents having suitable data and criteria; the speech did not report a deployed system or measured improvement.

Waller described two forms of agentic commerce, a topic explored in AI agents’ business applications. In agent-assisted shopping, an agent helps with product search while the consumer chooses and pays. In agent-delegated shopping, a consumer authorizes an agent to make purchases under specified constraints. He gave the example of an agent using a pre-funded virtual card to order groceries based on past purchases. Waller said consumer-to-business transactions could be the first area of adoption, with assisted uses preceding more autonomous ones.

At a glance
reportWhen: Speech delivered September 29, 2026; pu…
The developmentChristopher Waller outlined potential uses and risks of AI in payments and agentic commerce in a speech delivered at Sibos 2026.

Payment Security Meets Autonomous Shopping

AI could affect both the checks that allow payments to move and the decisions that initiate purchases. If screening systems reduce false positives, payment providers may be able to direct more attention to higher-risk cases. If routing tools improve cost or speed, cross-border services could benefit, although Waller offered no performance figures or implementation timeline.

Delegated commerce shifts part of the purchasing process from a person to software. That raises practical questions about how consumers set spending limits, approve purchases, and resolve errors. Waller’s distinction between assisted and delegated models matters because the latter gives an agent authority to act, making reliable controls central to consumer trust.

Products Worth Considering

FLEXISPOT Hako G2 Queen Bed Frame with Headboard, Solid Rubberwood, Tool-Free Japanese Joinery

FLEXISPOT Hako G2 Queen Bed Frame with Headboard, Solid Rubberwood, Tool-Free Japanese Joinery

LG UltraGear 34-Inch Curved QHD Gaming Monitor, 160Hz, FreeSync Premium

LG UltraGear 34-Inch Curved QHD Gaming Monitor, 160Hz, FreeSync Premium

As an affiliate, we earn on qualifying purchases.

From Payment Tools to Shopping Agents

Waller framed payments as an industry that has already adopted AI for operational tasks, including fraud detection and reconciliation. His speech at the annual Sibos conference expanded that discussion from tools that support payment staff to agents that could plan and execute multi-step transactions.

He said he had hosted an industry roundtable on agentic commerce at the Federal Reserve earlier in the summer. According to his account, participants generally viewed the market as being at an early stage, while recognizing that adoption at scale could reshape commerce and payments. His remarks set out possible directions and concerns, rather than announcing a Federal Reserve program or a new payments product.

“I expect LLMs to augment, rather than replace, faster, more traditional anomaly detection methods.”

— Christopher Waller, Federal Reserve governor

Adoption and Safeguards Remain Open

Waller did not specify when agent-delegated commerce might become common, which trust mechanisms would be required, or who would set and enforce them. The speech also did not identify particular products or payment providers already using agents to complete purchases at scale.

Although Waller cited research on reduced false-positive alerts, he did not provide the study details or a numerical estimate in the speech excerpt. The scale of any gains in cross-border payment efficiency is also unclear. He warned of AI-enabled cyber risks but gave no estimate of their likelihood or potential impact on payment networks.

Testing Controls as Use Expands

Waller’s speech did not set out a formal next milestone or policy timetable. The near-term development to watch is whether payment firms and AI providers move from assisted shopping and screening tools toward systems that can act on a consumer’s authorization.

As that work develops, the questions raised in the speech include how agents will receive real-time information, what limits consumers can set, and how providers will protect payment systems against attacks. Waller’s remarks indicate that any broader use of delegated commerce will depend on building trust mechanisms and safeguards, but he did not describe a specific framework.

Key Questions

What did Christopher Waller say about AI and payments?

He discussed potential uses of AI in cross-border payment screening and routing, along with risks tied to cybersecurity and AI agents making purchases.

What is agentic commerce?

It is the use of AI agents in shopping and payments. In an assisted model, the consumer makes the purchase decision and handles payment; in a delegated model, the consumer authorizes an agent to shop and pay within set constraints.

Did Waller say AI has already improved cross-border payment performance?

No specific improvement in speed, cost or reliability was reported. Waller described routing and currency-conversion optimization as potential uses for AI agents.

What risks did Waller identify?

He pointed to cybersecurity threats and the possibility of unintended purchases when an agent is authorized to shop autonomously. He said delegated commerce would require more extensive trust mechanisms and guardrails.

Source: primary

This content is for general information only and is not financial, tax or legal advice. Consult a qualified professional for decisions about your money.
You May Also Like

Stock market outlook: S&P 500 to lose much of 2026 gains as ‘speculation is hitting extreme levels’

Analysts predict the S&P 500 will decline significantly in 2026 as market speculation reaches extreme levels, raising concerns for investors.

Kimi K3’s Market Strategy: Leveraging AI To Accelerate And Stabilize

Moonshot AI’s Kimi K3, with 2.8 trillion parameters and priced at Western mid-tier levels, marks a major move in Chinese AI competitiveness.

The 2028 Model Lab Endgame: How Six Becomes Two, Three, or Twelve

A scenario forecast explores how six Western frontier AI labs may consolidate into two, three, or twelve by 2028, with significant implications for AI development and capital.

From Training To Talking: How AI Models Learn To Answer

An in-depth look at how AI language models develop capabilities, shape behavior, and generate responses without learning from conversations after deployment.