IRS risk chief warns AI payments outpace regulatory speed

Two years ago, AI was used for research and writing tasks. Soon after, it gained capabilities like coding, building apps quickly, automating workflows, and creating high-quality media. Now, AI systems are processing millions of payments.

Regulators have not kept pace. As AI agents gain more financial control, this disparity could become dangerous.

During the Future of Money summit at UN Headquarters in New York, BeInCrypto moderated a panel on this issue. Participants included Dottie Romo (IRS Chief Risk and Control Officer), Dino Cataldo Dell’Accio (UN Joint Staff Pension Fund), and Julius Moye (Mastercard).

AI Payments Already Taking Place at Scale

BeInCrypto's State of AI Agent Payments 2026 research illustrates current machine-driven payment activity.

From July 23 to August 26, researchers observed 6.4 million x402 payment transactions worth $119,947 on Base and Solana.

Most were tiny. 90.8% of transfers were under one cent, though the AI agent has processed nearly 200 million settlement transactions since launch, the research states.

Payments stay small because agents mostly pay for discrete digital resources like data or API access. The transaction volume highlights how machine commerce operates differently.

Regulators May Require Algorithmic Oversight

The IRS risk chief cautioned that traditional oversight might not handle that velocity.

“They’re making millions of decisions in minutes,” said Dottie Romo from the IRS.

Romo noted that regulators depend on periodic reports to detect fraud and control issues. Autonomous finance could render that method too slow.

Would algorithms need to watch algorithms? Romo indicated that some automated oversight would probably be required.

“We’re not going to be able to do that in a fast enough pace,” said Romo.

She advocated for increased real-time monitoring across markets while retaining human involvement in critical decisions.

Who Controls the Kill Switch?

Mastercard's Moye cited the 2012 Knight Capital incident and Terra/Luna collapse as warnings about insufficient safeguards in automated systems.

His framework involves machines identifying abnormal activity and automatically containing it, with severe incidents escalated to humans.

“It’s really using AI and machines to apply the tourniquet and stop the bleeding and then have humans come in to do the surgery,” said Julius Moye, Manager at Mastercard’s Financial Crime Solutions.

Dell’Accio contended that automation does not absolve responsibility. Accountability, he said, should be traceable via three queries:

Who developed the code? Who implemented the code and who oversees the code?”

A Dystopian Prospect

The unsettling truth is that AI now plays a prominent and assertive role in modern finance, conducting millions of transactions monthly.

Regulators cannot realistically bridge this gap. That speed is beyond human capability. Consequently, a future where machines monitor machines may be approaching.

This may sound dystopian, yet policymakers are already debating it.