A few months ago, I received an invitation to attend a lecture with the mind-numbing title, “The Transformation of Tax & Customs Law into the Digital World of Processes, Automation and Predictions.” I wasn’t able to attend, but it reminded me that like the android hosts in Westworld, the IRS was developing an artificial consciousness.
Or, as two lawyers writing about the increasing use of artificial intelligence and big data in tax enforcement and compliance asked, “Do IRS computers dream about tax cheats?”
We don’t believe that the IRS as an agency is suddenly becoming sentient. But for more than a decade, the agency has been using data and analytics to parse tax return data to uncover taxpayer errors, and occasionally, tax fraud. For instance, in the decade from 2007-2017, the volume of data analyzed by the agency grew by a factor of 100.
The information technology infrastructure necessary to analyze this data has evolved as well, making initiatives such as Operation Hidden Treasure possible. In an effort to root out tax evasion involving cryptocurrency investments, the agency is working with blockchain analytics firms to develop “signatures” or potential indicators of tax fraud, by “analyzing blockchain and de-anonymizing [crypto] transactions.” The objective is to “track, find, and work to seize crypto in both a civil and a criminal setting.”
But that’s really just the tip of the iceberg. If you’ve experienced an IRS audit lately, you might have noticed the agency is demanding a lot more information than it once did. For instance, one firm that represents taxpayers in cryptocurrency audits publishes this list of information the IRS always requests in such audits. It’s far too detailed to publish here but suffice it to say that the IRS is demanding detailed records for the acquisition and disposition of any cryptocurrency, including emails, text messages, tweets, counterparties, borrowing, lending, and much more. All this data is analyzed by the agency’s increasingly sophisticated AI tools.
Keep in mind Operation Hidden Treasures isn’t the IRS’s only “campaign” in which it’s leveraging big data and AI to uncover tax evasion and fraud. It’s using similar tools to identify taxpayer noncompliance in dozens of different subjects. Other than cryptocurrencies, the foreign activities of taxpayers are of particular interest to the IRS. Thus, legitimate tax reduction techniques we have long recommended to our clients are in the agency’s crosshairs, including the foreign earned income exclusion and expatriation.
Among the most significant investments the IRS has made in this area is an artificial intelligence tool called Lead and Case Analytics (LCA), developed in conjunction with Palantir Technologies, a company focusing on data analytics. The IRS justifies the approximate $100 million cost to develop LCA on its contention that, “Today’s sophisticated financial schemes to defraud the government demand the technology to compile disparate case data and the analytical tools to wade through complex financial records to identify fraudulent activity.”
Thus, even though the official IRS audit rate has fallen sharply in recent years, every tax return is subjected to computer scoring and compared to returns of people with similar income and occupations. That means the effective audit rate is 100%.
The goal, as always, is to reduce the “tax gap,” which IRS Commissioner Charles Rettig told Congress last year could be as large as $1 trillion per year. This is the difference between what the IRS thinks taxpayers should be paying and what it collects.
Frankly, we find this enormous number preposterous. To the extent that the tax gap exists at all, it’s mainly because of the insane complexity of the Tax Code. Nina Olson, the IRS’s former taxpayer advocate, told Congress in 2006 that in 94% of audits, IRS examiners found no willful tax evasion – only inadvertent errors by taxpayers and their advisers who simply don’t understand the monstrously complex rules. Instead of worrying about a tax gap that is largely the result of a Tax Code no one understands, why not simplify the rules?
Sadly, that seems unlikely to occur, since politicians of both parties are only too eager to mold the Tax Code to fit the needs of their constituents. Which leads us to ponder: can individuals and small businesses use AI tools themselves to fend off the IRS?
The answer, perhaps surprisingly, is “yes.” While AI tools are mainly being used by tax administrators and Fortune 500 companies, they’re becoming more accessible all the time for everyday users, with no coding experience required. The goal is to encourage a settlement with minimum or no additional tax being due.
AI analysis is inherently probabilistic. When you run an AI simulation, the outcome will be an answer that, based on the data loaded into the system, is expressed as a probability. Once these tools are more widely deployed, if you or your business is audited by the IRS, and there is a disagreement on your tax returns position, you and your advisors can tell the agency:
“Don’t take our word for it. Using the same data you have, we analyzed it independently. Our AI program is telling us that we have a 95% probability of winning this case. Are you still sure you don’t want to settle?”
The bottom line is that despite the IRS hype about the tax gap and its use of big data and AI, there’s nothing stopping the rest of us from using the same tools as they become available.
