The IRS is staffing up with investigators to go after all sorts of criminal activity, but criminals are using artificial intelligence tools to launch more sophisticated fraud schemes, and in greater volumes.

The agency’s criminal investigation branch, however, is also relying on AI tools to stay one step ahead of fraudsters.

Jarod Koopman, IRS Criminal Investigations’ executive director of cyber and forensics, said in a recent interview that online payment fraud now exceeds $360 billion annually, and that check fraud is “skyrocketing.”

“What used to take a significant amount of effort, going into some type of a social media-type exploit or a hack, they can now do this with AI that’s much more efficient, much more effective, and certainly much more volume at high speed,” Koopman said.

The IRS, under the Inflation Reduction Act, recovered more than $1.3 billion from about 1,600 millionaires who had not paid overdue tax debts or filed tax returns in recent years.

Koopman said the agency isn’t currently using AI in these tax recovery cases, but IRS-CI is reviewing AI as a use case for “that exact scenario.”

“They want to make sure what they’re doing is implementing in a way that is very unbiased and done in a way that’s governed and utilized to ensure both the privacy, but also the integrity of the application of the tax law,” Koopman said.

IRS enforcement operations, more broadly, are trying to keep up with emerging technology’s impact on criminal financial activity. AI tools are accelerating the development of deepfake schemes, as well as misinformation and disinformation campaigns that led to fraud.

Meanwhile, a trillion-dollar global cryptocurrency market gives fraudsters more opportunities to hide criminal sources of revenue from federal law enforcement agencies.

“From a general standpoint, we’ve seen not only fraud continue on the rise, but more sophisticated frauds in the way of AI, cyber components, in addition to traditional financial fraud,” Koopman said.

He estimates that about 50% to 60% of IRS-CI’s casework relates to taxes. IRS-CI also investigates a wider scope of criminal activity with a financial component — including public corruption, terrorism funding, organized crime and drug trafficking and money laundering.

A recent IRS-CI investigation of child exploitation material, paid for through cryptocurrency transactions, led to 300 arrests worldwide.

IRS-CI is the sixth-largest law enforcement agency in the federal government, but lacks the robust staffing of larger entities such as the FBI and Immigration and Customs Enforcement’s Homeland Security Investigations (HSI).

“Some of these agencies have 30,000-40,000 employees, versus us coming in at 3,500. We’re also a very small portion of the IRS as a whole,” Koopman said. “We tend to be an agency that’s very flexible and nimble. We’re small enough that we can drive innovation. And that’s really helped us, I think, in the long term.”

IRS-CI is hiring law enforcement personnel with expertise in international banking, anti-money laundering, cybercrime and cryptocurrency. Nearly all its hires, however, are required to have a background in accounting.

Koopman said IRS-CI in recent years had the resources to bring its staffing up to levels not seen since previous decades. In recent years, it’s grown from about 2,700 employees to about 3,500. About 2,500 of them are frontline special agents, and the rest are professional staff.

“Over the last couple of years, we’ve actually been able to increase our employees on roles a bit to help continue to meet the demands,” he said. “This is just basically getting us back to staffing levels that were back in the 90s and the early 2000s. So nothing of significant growth, but definitely something that’s been helping.”

The IRS is generally using AI to make more effective use of the data already available to its workforce. This includes operational data from previous and ongoing casework, as well as third-party data from other financial institutions and open-source intelligence.

“All of that data is pretty vast and it’s large volumes of data. AI has certainly made us much more effective and efficient in sifting through that to find patterns, to identify fraud methodologies or typologies, to identify current threats or vectors that might be red flags of issues that we were not aware of prior,” Koopman said.

Koopman said that AI, above all, is helping the agency flag and detect patterns in the data it has, then have employees investigate further.

“It’s like having AI and large language models be able to match up against our data internally to be able to give us the results and the outputs that we’re looking for to make decisions, not for the AI to use the data to make decisions directly,” he said.

Koopman said the IRS is also relying on new privacy enhancing technologies that allow the agency to share sensitive information with law enforcement partners and third-party financial institutions through encrypted channels, and allow those organizations to run analytics or other models against those data sets without ever having access to the source data.

“If we want to correspond with a third party, we have to provide them with some type of a name or a Social [Security number] or something or handle to be able to get information back. There’s a trust factor there and there’s also a risk tolerance that we have to provide something in order to get the information that we wouldn’t readily have available otherwise,” Koopman said.

“That’s really a game changer when you start talking about some of these capabilities that we’re looking to do internally, because it just further enhances the security of the data that the government has and that all of these other partners have and not constantly shifting around information that’s pretty valuable to criminals and others,” he added

Copyright © 2025 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.