This tax season is different from others because of a loss of approximately 25 percent of the IRS workforce in the last year. With a budget boost in 2023, the IRS began to modernize its technology and hire much-needed talent to staff, initiate, and complete more specialized audits. Since then, the strategy has changed significantly. The workforce has dramatically thinned, and priorities have veered away from high-asset tax liability collection. That said, it is important to understand how AI will enhance enforcement activities at the IRS.
In the past, the IRS has relied on numerous strategies to detect civil tax issues and criminal tax evasion through random selection, selection by association with someone already audited, error flags, and underreporting triggers. In recent years, the IRS has used algorithms to select tax returns with concerning characteristics. The introduction of Large Language Models (LLMS) enhances and transforms these efforts.
In late 2025, the IRS discussed its use of AI. In addition to operations, procurement and customer service, the agency is using AI for enforcement and detection to deeply amplify its capabilities in a way that impacts:
Time: While audit and IRS Criminal Investigations (C-I) will always be major enforcement tools, the speed at which AI can analyze data for patterns, irregularities, mistakes, and other inconsistencies will increase the rate of investigation and the sheer number of issues detected.
Deeper analysis, not just detection: While earlier automation tools could flag specific information or errors for review by an analyst, LLMs can review data sets line by line to provide more meaningful results.
Personal spending vs income: The IRS Automated Underreporter (AUR) is a partially automated tool used to match tax returns with information already gathered by the IRS. A useful tool, the AUR helps identify taxpayer data compared to what has been reported by employers, vendors, and the like. LLMs can do even more by matching reported income to spending patterns and asset ownership to identify inconsistencies in what is reported and what is spent. For example, a disparity between income and assets and the taxpayer's personal and corporate returns could trigger IRS interest in payroll tax or other types of fraud.
Every audit helps: Because the IRS is using LLMs, the tools will improve with greater exposure to data that leads to IRS audits. Those involved in tax crimes rely on remaining hidden. Machine learning will take away some of the opacity on which tax evasion depends.
For taxpayers and business owners with something to hide from the IRS, the future is going to be problematic. To better understand if your actions are noncompliant, read our ebook, “Understanding Tax Fraud.” If understanding AI better has helped you identify potential compliance risks, it may be time to evaluate your tax position more closely.
If you receive a worrisome notice of deficiency or IRS audit letter, speak with our tax group. At Robert J. Fedor, Esq., L.L.C., we provide experienced legal counsel on options to help you navigate the interest of the IRS in your return or business. Contact us or schedule a consultation by calling 440-250-9709. We serve clients internationally and in Northeast Ohio, Chicago, and New York City from our offices in Cleveland and Chicago.