The I.R.S. is using AI to Crack Down on Tax Evasion

Jossie Ward

Associate Editor

Loyola University Chicago School of Law, JD 2025


The Internal Revenue Service (I.R.S.) issued a press release on September 8, 2023, detailing how the agency plans to use at least part of the $80 million dollar allocation it received from the Inflation Reduction Act last year. I.R.S. Commissioner Danny Werfel plans to use the funds to make compliance enforcement efforts and tax evasion identification more effective and efficient. How does he plan to do this? The overwhelmed and perhaps overworked agency will be using artificial intelligence (AI) programs and features to expedite and assist with redundant processes as well as to audit parties that are too complicated or large for the I.R.S.’s current capabilities.

How the I.R.S. uses AI generally

The I.R.S. will use their AI programs to improve everything from detecting fraud and enhancing transparency to automating the tax return processing and providing taxpayer services. The I.R.S. has already began implementing AI chatbot services and tools to help taxpayers with their questions. To improve compliance, the I.R.S. is using AI to develop risk assessment models that help the I.R.S. identify specific taxpayers that are more likely to be non-compliant with federal tax laws than others. The I.R.S. is using a Risk-Based Collection Model to identify patterns in tax data that indicate fraud, and the model has thus far helped the I.R.S. recover billions in unpaid tax dollars. The Modernized e-File (MeF) system uses optical character recognition and AI to extract data from paper tax returns to automate the processing of paper tax returns. This means the I.R.S. is able to reduce its backlog of unprocessed returns.

Who is the I.R.S. investigating?

The I.R.S. has made clear the agency has a goal to prioritize examining higher earning individuals and entities. The I.R.S. has stated many times that even with increased audit capability, the agency does not intend to increase audits for Americans making less than $400,000 annually. However, the I.R.S. does intend to prioritize collection efforts on individuals earning more than $1 million annually that have a federal tax liability in excess of $250,000. In the past, the I.R.S. has examined less than 0.3% of all partnership returns filed. With the influx of staff and resources, the I.R.S. will finally be able to evaluate more partnerships for compliance. Moreover, the integration of AI has allowed the I.R.S. to initiate examinations of large partnerships like hedge funds, publicly traded partnerships, private equity groups, real estate investors, and major law firms, with 75 of the United States’ largest partnerships at the top of the list. Most of these partnerships, on average, have more than $10 billion in assets. Moreover, the I.R.S. has increased its scrutiny of partnership returns with assets of at least $10 million when the returns have substantial balance sheet discrepancies and insufficient explanation.

Why are partnerships so hard to handle?

The I.R.S. has struggled to keep up with big business for years, without the staff or resources to be able to address partnerships. From 2002 through 2019, the number of partnerships with over $100 million in assets and 100 or more partners has increased almost 600% while the audit rate for partnerships of this size has dropped to less than 0.5%. However, AI will allow the I.R.S. to dedicate manpower and more thoroughly investigate more complicated and nuanced cases they have been unable to handle before with their workforce. Specifically, this effort to ensure compliance will focus on high-income earners, large partnerships, and large corporations. Investigating partnerships aligns with the I.R.S.’s overarching efforts to investigate wealthier taxpayers in 2024, and the use of AI allows the I.R.S. to look into harder to evaluate assets like foreign bank accounts and digital assets like cryptocurrencies. The I.R.S. has found that 75% of taxpayers who use digital assets for exchanges do not comply with federal tax laws.

AI and the I.R.S. practical effects for the U.S.

Every year, the I.R.S. includes a projected return on investment (ROI) as a part of its budget justification. The ROI is projected to increase as their staff becomes trained and more productive in their roles with the ROI ranging from 5-9 or rather for every $1 increase in spending on the I.R.S.’s enforcement results in $5 to $9 of increased revenues. In the Budget in Brief for Fiscal Year 2024, the IRS reported that its previous ROI was for Fiscal Year 2022 was $6 in revenue for every dollar spent on enforcement efforts. When the I.R.S. was overwhelmed with unprocessed returns and unable to handle the cumbersome cases of partnerships, the ROI was still $6 for every dollar invested. The increased budget and implementation of AI technology will allow the I.R.S. to process more returns faster, expedite examinations, and help proliferate the process of identifying potential tax evasion or fraud. While the programs are new and there is still a risk of machine error, the potential for increased compliance and revenue makes the inclusion of AI technology a no-brainer.