Could Artificial Intelligence (AI) increase the likelihood of an IRS audit in your future? How does the IRS use AI to identify US taxpayers who attempt to hide assets, under-report income or otherwise cheat the IRS?

In a recent interview, IRS Commissioner Danny Werfel began to lay out a vision of where the IRS is headed and how AI will help the agency identify and target those who abuse the law to avoid paying taxes. The IRS has been “underfunded and understaffed” for too many years and AI is a cost-efficient and effective tool for identifying audit targets. Mr. Werfel specifically mentioned those who owe more than $250,000 in taxes, high-income taxpayers, and the Large Partnership Compliance (LPC) program.

The IRS has always worked to identify “patterns” of behavior and common strategies taxpayers use to hide income and assets, especially in offshore banks and entities. Implementing the Foreign Account Tax Compliance Act or FATCA (passed as part of the HIRE act and enacted in 2010) required offshore banks, investment houses, and even sovereign tax authorities to provide electronic information about US taxpayers directly to the IRS. “If you want to do business with or within the United States you will fully comply with the requirements of FATCA.”

In essence, FATCA requires all “Foreign Financial Institutions (FFI)” to provide detailed information to the IRS about the accounts and transactions of US taxpayers and foreign entities, such as corporations or trusts, in which a US taxpayer may hold an interest.

Swedish banks were the first to be targeted, and since 2014 the US Treasury and the IRS have successfully sued to penalize those institutions and sovereign tax authorities who refuse to comply. These enforcement actions continue to this day.

On December 4 (2023) the US Department of Justice issued a press release regarding a deferred prosecution agreement and a fine of $122.9 Million for “Swiss private bank Banque Pictet et Cie SA” who “admitted today to conspiring with US taxpayers and others to hide more than $5.6 Billion in 1,637 secret bank accounts in Switzerland and elsewhere and to conceal the income generated in those accounts from the IRS … Rooting out financial malfeasance remains a priority for this Office, and we encourage companies and financial institutions to come to us to report wrongdoing before we come to you.”

What do you think the agency will be doing with the information regarding the “1,637 secret bank accounts” and individual US taxpayers who hold any interest in them?

What are some specific examples? How does the IRS use AI to identify tax cheats?

How does the IRS use AI to identify tax cheats, and what are some specific examples of how AI will transform the IRS? The IRS now receives a substantial trove of electronic data from FFIs and sovereign tax authorities worldwide. Until recently, a lot of this information simply couldn’t be investigated. The agency could draw on this information when dealing with a targeted US taxpayer once they had been identified. Now, the IRS can let new, state-of-the-art computer systems and Artificial Intelligence do the work for them. AI is the game changer.

AI is now used to comb through all of the electronic information coming into the agency from around the world. In essence, AI simply needs to compare the information provided to the IRS with the tax filings of associated US taxpayers. If the information reported to the IRS doesn’t match the information provided by the US taxpayer AI will “raise a red flag” and begin the audit process.

AI is able to help identify patterns of behavior designed to avoid tax reporting to the IRS and to investigate specific groups of US taxpayers. Commissioner Werfel raised the example of “large partnerships” which, to date, represent some of the most “complex cases the IRS faces.” AI can quickly help to pierce through complex financial data, tax filings, offshore data and other sources of information that the agency previously didn’t have the resources to undertake. Large partnerships represent a “wide range of industries and activities where it's been far too easy for tax evaders to cut corners. The IRS has simply not had enough resources and staffing to address partnerships," said Mr. Werfel.

Artificial Intelligence is able to comb through all offshore reporting flowing into the IRS. Commissioner Werfel noted examples of failure to fully and accurately provide information to the IRS through FinCEN Form 114, commonly referred to as the “FBAR” (Report of Foreign Bank and Financial Accounts). US taxpayers are required to disclose every foreign bank or investment account they hold if the aggregate total of all of those accounts exceeds $10,000 at any point in a calendar year, even for one day or one hour. AI has identified “hundreds of US taxpayers” who failed to file an FBAR on aggregated account balances which exceed $1 Million. These “most egregious potential non-filer FBAR cases” are immediately targeted for audit.

A recent AI review of the IRS Virtual Currency Compliance Campaign identified that as much as 75% of US taxpayers identified through the information now provided to the agency by digital currency exchanges were not compliant with US tax laws regarding the disclosure of cryptocurrency and NFTs on their tax returns. AI makes it easy to target and audit them.

AI has helped the IRS to identify tax schemes involving the fictitious “brokerage” of labor. In this scheme, a labor contractor makes IRS Form 1099 payments to their “subcontractors.” However, the subcontractors are not real or genuine and have no relationship whatsoever to the actual business of the labor contractor. The subcontractors are actually a shell corporation that funnels the money back to the corporation of the labor contractor. The resulting substantial tax deductions claimed by the labor contractors are being investigated for civil and criminal tax evasion.

AI is a Game Changer for the IRS

How does the IRS use AI and how will this make a difference to the agency in the future? IRS Commissioner Werfel reflected on how past cuts at the IRS had "enabled wealthy individuals, large partnerships, complex and large corporations to come up with increasingly creative ways of reaching their most tax-advantaged status. The issue is that in many of those cases they took steps that are technically tax evasion under the tax law, and that's where this focus is. We now have the ability to invest in analytic capacity, people capacity and technology capacity to catch up from 10 years of under-investment and to establish a new expectation across our wealthy filers that we have the ability now to see where income is being shielded where it should not be shielded."

The National Bureau of Economic Research recently revised their study “Tax Evasion at the Top of the Income Distribution: Theory and Evidence” (© 2021 J. Guyton, P. Langetieg, D. Reck, M. Risch, and G. Zucman) which concluded the top 1% of US wage earners fail to report more than 20% of their actual earnings to the IRS.

That is about to change. Artificial intelligence is one of the best, and most efficient and effective tools available to the IRS. AI is a game changer.