IRS Artificial Intelligence and Tax Audit Risk: Overview for U.S. Taxpayers
Artificial intelligence has moved to the center of how the Internal Revenue Service selects tax returns for audit, detects tax fraud, and administers the U.S. tax system. According to a March 2026 report by the Government Accountability Office, the IRS maintained 126 active AI use cases as of June 2025 — up dramatically from just 10 in August 2022 — covering audit selection, fraud detection, income matching, taxpayer services, and enforcement analytics.
At the same time, the IRS workforce fell from over 100,000 employees at the end of 2024 to approximately 75,000 by mid-2025, a reduction of roughly 25,000 positions driven by the Department of Government Efficiency (DOGE) layoffs, probationary terminations, and voluntary buyouts. Reports indicate that plans to cut the workforce by as much as half are under active consideration.
The result is an agency that is simultaneously deploying more artificial intelligence and exercising less human oversight — a combination that has significant practical consequences for every U.S. taxpayer.
For individuals, small business owners, high-net-worth taxpayers, large partnerships, cryptocurrency holders, and taxpayers with offshore assets, understanding how IRS artificial intelligence operates, what it targets, and what rights remain available is no longer optional — it is a foundational component of sound tax compliance. Experienced U.S. tax lawyers are already seeing the real-world impact of AI-driven IRS enforcement in the matters they handle, and the time to understand this landscape is before an IRS audit notice arrives, not after.
IRS AI and Tax Enforcement Technology: Background and Historical Context
The IRS has used automated screening tools to select tax returns for audit for decades. Its legacy Discriminant Function System (DIF) — a statistical scoring algorithm that assigns each filed return a numeric score based on the likelihood of unreported income or overclaimed deductions — has long been the primary mechanism for IRS audit selection. AI and machine learning have not replaced the DIF but have been integrated into and layered on top of it, substantially expanding its reach and sophistication.
The IRS now uses AI in both the selection of tax returns for audit and to conduct the audit itself. It identifies high-risk returns filed by large corporations, complex partnerships, high-net-worth individuals, and users of digital assets. According to the GAO, the IRS had 126 active AI use cases as of June 2025, up from just 10 in August 2022, with 61% of those use cases still in development. The IRS categorized most use cases as either improving operational efficiency or enhancing tax compliance and fraud detection.
The expansion of AI use cases has accelerated precisely as the agency’s human workforce has contracted. In 2025, the IRS lost approximately 25,000 employees — roughly one quarter of its peak workforce of over 100,000 — through a combination of DOGE-directed layoffs, probationary terminations, and voluntary departures. Officials in the IRS’s Research, Applied Analytics and Statistics group, which supports AI development, oversight, and use, confirmed to the GAO that they lost 63 employees who had been working full or part-time on AI. The agency is losing the very personnel it needs to build, validate, and oversee the AI tools it is simultaneously deploying at unprecedented scale.
The stakes are considerable. With an estimated $688 billion annual tax gap, the IRS is under significant pressure to make enforcement more effective. AI is the primary mechanism through which the agency is attempting to close that gap with a diminished workforce — a structural shift that makes understanding AI-driven enforcement essential for every U.S. taxpayer and their advisors.
How IRS AI Selects U.S. Tax Returns for Audit: Key Issues and Findings
AI-Powered Audit Selection: From DIF Scores to Machine Learning
The IRS’s AI-driven audit selection operates across multiple programs and divisions. Machine learning models now analyze millions of tax returns simultaneously, scoring them for IRS audit potential. AI models are used to help select a representative sample of taxpayer returns for audit and to identify returns that are more likely to have errors or to owe additional taxes.
According to the GAO, the IRS is using AI more broadly to review large volumes of tax and other data and to identify returns at higher risk of noncompliance. For returns already under examination, IRS AI tools may supplement the examining agent’s analysis. For open years not yet under audit, the GAO confirmed that the IRS uses AI tools to identify noncompliance issues in filed returns on an ongoing basis — not only at the point of filing. Taxpayers should not assume that the scope of IRS inquiry will remain limited to initially identified issues.
The Large Partnership Compliance program, launched in 2021, uses machine learning to assess accounting rules and tax law compliance, enabling more efficient audits of complex entities. AI also compares taxpayer-reported income with third-party data — such as bank records and property transactions — to issue CP-2000 notices for discrepancies, and identifies outliers within specific industries, enhancing the traditional DIF scores used for audit selection. The Large Business and International Division has been using an AI model that screens each large partnership’s tax documents to determine the risk of incorrect reporting. As of April 2024, this model had selected 82 returns for audit for the 2021 tax year, though there was not yet data demonstrating whether those selections were effective. The IRS is notably secretive regarding the development of its AI programs.
IRS AI and Workforce Reductions: A Diminished Human Oversight Layer
The intersection of aggressive AI deployment and dramatic workforce reduction creates a risk that is unique to the current IRS environment and distinguishes it materially from any prior period of IRS enforcement. Former National Taxpayer Advocate Nina Olson warned that IRS workforce reductions will hinder tax collections, discourage voluntary compliance, and erode taxpayer trust, arguing that there needs to be a trained human being who understands the challenges taxpayers are experiencing, and that without skilled agents with the training and resources to address sophisticated tax matters, the whole tax system becomes more unfair.
Former IRS Commissioner Charles Rettig stated that automation can leverage human resources but not replace them, and that the IRS’s approach should have been to use AI to make itself better and more efficient — not to reduce the workforce. The current trajectory departs from that model, with AI increasingly filling functions previously performed by trained human agents. Concerns have been raised about whether there are sufficient human resources to review AI conclusions, whether remaining employees may be reluctant to challenge the outputs of automated systems, and whether there are sufficient safeguards as the IRS uses external databases and contractors to assist with AI applications.
Algorithmic Bias: A Documented and Growing Concern
One of the most significant criticisms of IRS AI is the documented risk of algorithmic bias producing racially disparate audit outcomes. Independent studies show that Black taxpayers are audited at rates three to five times higher than others. The GAO identified unintentional algorithmic biases as a potential cause, noting that when AI trains on historical data containing existing biases, it perpetuates past discrimination through automated systems. With workforce cuts leaving insufficient human oversight, the risk of unfair outcomes grows.
As explored by both the GAO and the Treasury Inspector General for Tax Administration (TIGTA), AI programs are created using pre-existing data, and to the extent that data has been impacted by prior biases and social inequities, the resulting AI program may continue to perpetuate those disparities. This is not a theoretical concern — it has been independently documented and reported to Congress, yet the IRS continues to expand its AI footprint without having fully resolved it.
Lack of Transparency and GAO Documentation Concerns
The IRS’s opacity about its AI systems compounds the fairness and accountability concerns. The GAO found that the IRS’s AI inventory was incomplete — it did not include all the ways AI was being used, did not identify how tools would benefit the agency, and 65 of the 126 use cases were either too sensitive for public reporting or were research and development efforts exempt from disclosure. The GAO has called for better documentation and transparency around the IRS’s use of AI, and advisory panels have recommended improvements, but taxpayers remain largely in the dark about the algorithms making decisions about their returns. Taxpayers selected for audit are not told whether it was humans or AI that flagged their return.
The inability to explain audit selection is not merely an administrative shortcoming — it has direct consequences for taxpayers who wish to understand and contest the basis on which they have been targeted, and for experienced U.S. tax lawyers seeking to mount an effective defense.
IRS AI and Cryptocurrency: Elevated Enforcement Risk
Cryptocurrency and digital asset holders represent one of the IRS’s most actively targeted enforcement categories. AI is used specifically to identify high-risk returns involving users of digital assets. The IRS has access to third-party reporting data from exchanges operating in the United States, and AI cross-referencing of that data against filed returns is a standard enforcement tool. U.S. taxpayers who have failed to report cryptocurrency income, capital gains, or dispositions — including those holding foreign cryptocurrency accounts with FBAR reporting obligations — face a materially elevated risk of IRS identification and examination.
The “Robodebt” Warning: A Cautionary Tale for AI-Driven Tax Enforcement
Former National Taxpayer Advocate Nina Olson highlighted Australia’s Robodebt debt recovery system as a cautionary tale for AI-driven tax enforcement. That system used automated data matching in a catastrophic fashion, wrongfully sending debt notices to welfare recipients at scale. The Australian government ultimately paid hundreds of millions of dollars in settlements after the program was found to be unlawful. The warning is directly applicable to IRS AI audit selection: automated systems operating with minimal human oversight, trained on imperfect data, and designed to maximize enforcement yield can produce systematic errors at scale — with real and serious consequences for individual taxpayers who have done nothing wrong.
IRS AI Tax Enforcement: Implications for U.S. Taxpayers and Businesses
The IRS’s deployment of AI for audit selection and tax enforcement has several important practical implications for U.S. taxpayers and their advisors:
- Audit selection is now substantially algorithmic, with less human review than at any prior point. With 126 active AI use cases and a workforce that has fallen from over 100,000 to approximately 75,000 employees, the proportion of audit selection decisions effectively driven by automated scoring has never been higher. Discrepancies between filed returns and third-party data are identified and flagged algorithmically before any human agent reviews the file. Accurate and complete reporting reconciled against all available third-party information is more important than ever.
- AI tools are not limited to audit selection — they operate throughout the examination process. The GAO has confirmed that IRS AI tools supplement the examining agent’s analysis for returns already under audit, and continue scanning open years not yet under examination. Taxpayers should not assume that the scope of IRS inquiry will remain limited to initially identified issues.
- Large partnerships and complex business structures face targeted AI scrutiny. The Large Partnership Compliance Return Selection Model specifically screens partnership returns for compliance risk. High-net-worth individuals with interests in complex partnerships, tiered structures, or offshore entities are at elevated audit risk under the current AI enforcement regime.
- Cryptocurrency and digital asset holders face active AI-driven enforcement. Third-party exchange data is routinely cross-referenced against filed returns by IRS AI tools. U.S. taxpayers with unreported cryptocurrency income, undisclosed foreign digital asset accounts, or unresolved FBAR obligations face a meaningfully elevated risk of IRS identification.
- The reduced human oversight layer increases the risk of erroneous audit selection without practical remedy. A taxpayer flagged by an algorithm trained on biased historical data, or that mischaracterizes a legitimate transaction as a compliance anomaly, faces an audit that may proceed without any human agent having meaningfully evaluated the underlying basis for selection. Challenging the algorithmic basis of an IRS audit is not currently a straightforward procedural option. The practical consequence is that a taxpayer flagged in error has no front-end remedy — defense must be mounted after the examination begins, at the taxpayer’s expense.
- Earned Income Tax Credit claimants face documented disproportionate algorithmic scrutiny. The independently documented bias in IRS AI tools means that EITC claimants and certain demographic groups face audit rates the GAO has identified as disproportionate — an ongoing area of concern that experienced U.S. tax lawyers and taxpayer rights advocates are pressing the IRS to address.
Pro Tax Tips for Taxpayers Facing IRS AI-Driven Audit Risk
Experienced U.S. tax lawyers recommend that taxpayers and their advisors treat the IRS’s AI enforcement environment as a standing compliance obligation — not a reactive concern triggered only upon receipt of an IRS audit notice. Every source of income, including cryptocurrency dispositions, platform-economy earnings, foreign investment income, and partnership distributions, should be fully and accurately reported and reconciled against all available third-party information returns. IRS machine learning tools are designed to detect outliers relative to industry norms, peer groups, and historical filing patterns; even legitimate deviations benefit from clear contemporaneous documentation explaining the underlying facts.
For U.S. taxpayers with unreported offshore assets, undisclosed foreign accounts, or unfiled FBAR or FATCA reporting obligations, voluntary disclosure options remain available but are more limited than in prior years. The IRS Offshore Voluntary Disclosure Program was permanently closed in September 2018. Current options for non-willful taxpayers include the Streamlined Filing Compliance Procedures — either the Streamlined Domestic Offshore Procedures or the Streamlined Foreign Offshore Procedures depending on residency — which provide penalty relief for taxpayers who can certify that their non-compliance was non-willful.
For taxpayers whose conduct may be characterized as willful, the IRS Criminal Investigation Voluntary Disclosure Practice remains available and provides an opportunity to come forward before the IRS initiates contact, with the possibility of avoiding criminal prosecution, though it does not guarantee immunity and penalty exposure can be significant. An experienced U.S. tax lawyer should be consulted before any voluntary disclosure submission is made, as the choice of program, the completeness of the submission, and the characterization of the taxpayer’s conduct are all critical and consequential determinations.
Taxpayers who receive an IRS audit notice or examination letter should engage an experienced U.S. tax lawyer immediately and before providing any response or documentation to the IRS. AI-assisted audits may proceed from algorithmic assumptions rather than a full understanding of the taxpayer’s actual facts and circumstances. Effective representation by one of the best tax lawyers requires identifying what triggered the examination, organizing a response that addresses the underlying IRS concern directly and completely, and ensuring that the taxpayer’s rights under the Taxpayer Bill of Rights — including the right to be informed, the right to challenge the IRS’s position, and the right to retain representation — are fully preserved throughout the process.
Frequently Asked Questions About IRS AI and U.S. Tax Audits
Does the IRS use AI to decide whether to audit a specific taxpayer?
Yes, AI and machine learning tools play a significant and growing role in IRS audit selection. The IRS uses AI to score filed returns for audit potential, cross-reference reported income against third-party data, flag outliers relative to industry and peer group norms, and screen complex partnership and corporate returns for compliance risk. The IRS maintains that human agents make final audit decisions, but with a workforce that has fallen from over 100,000 to approximately 75,000 employees and 126 active AI use cases confirmed by the GAO, the practical weight of AI scoring in determining which files proceed to examination has never been greater.
Can a taxpayer find out whether AI was used in selecting their return for audit?
The IRS has been criticized by the GAO for failing to adequately document its AI models and for maintaining a significant number of AI use cases that are not subject to public reporting. There is currently no formal procedural mechanism through which a taxpayer can readily discover the specific AI tools or scoring methodologies that led to their audit selection. The GAO has explicitly noted that taxpayers selected for audit are not told whether it was humans or AI that flagged their return. This is an area where experienced U.S. tax lawyers and taxpayer rights advocates are actively seeking greater transparency and accountability from the IRS.
What types of income or transactions does IRS AI most commonly flag for examination?
IRS AI tools are particularly active in identifying unreported or underreported cryptocurrency transactions and digital asset dispositions, discrepancies between reported income and third-party information returns, outlier deductions or losses relative to industry norms, complex partnership structures with potential compliance risk, offshore accounts and foreign asset reporting obligations, and EITC claims flagged as potentially erroneous. Business owners, investors, high-net-worth individuals, and cryptocurrency holders are at elevated risk of AI-triggered examination.
Is the IRS’s use of AI in tax enforcement legal?
Yes. The IRS has broad statutory authority to select returns for examination and to use data analytics and automated tools in its enforcement processes. However, the GAO has documented significant transparency, documentation, and workforce adequacy shortcomings in IRS AI programs, and the documented racial disparities in algorithmically selected EITC audits raise serious fairness and accountability questions that experienced U.S. tax lawyers and civil rights advocates are actively examining. The IRS is subject to the Taxpayer Bill of Rights, the Privacy Act of 1974, and applicable constitutional due process requirements, and the extent to which current AI practices fully satisfy all of those obligations in every instance remains a contested question.
What should I do if I receive an IRS audit notice?
Engage an experienced U.S. tax lawyer immediately and before responding to the IRS in any form. An experienced U.S. tax lawyer can assess the scope and basis of the examination, advise on your rights and options, organize your documentation, and represent you in dealings with the IRS in a manner designed to protect your legal position and achieve the best possible outcome.
Key Takeaways: IRS Artificial Intelligence and U.S. Tax Audit Risk
The IRS’s accelerating deployment of artificial intelligence in audit selection and tax enforcement — combined with a dramatic reduction in the human workforce available to review and check AI outputs — represents the most significant structural shift in U.S. tax administration in decades. With 126 active AI use cases confirmed by the GAO as of June 2025, machine learning embedded in the DIF scoring system, dedicated AI models targeting large partnerships and high-net-worth individuals, and aggressive cross-referencing of cryptocurrency and third-party data, the IRS is now an AI-driven enforcement agency whose workforce has fallen from over 100,000 to approximately 75,000 employees — with further reductions under active consideration.
The documented concerns are serious: algorithmic bias producing racially disparate audit outcomes, inadequate documentation of AI model specifications, opacity about audit selection methodology, and insufficient human review capacity to check AI conclusions. These are not theoretical risks — they have been identified by the GAO, TIGTA, and the former National Taxpayer Advocate, and they directly affect real taxpayers who may find themselves in IRS examinations triggered by erroneous or biased algorithmic scoring with limited ability to discover or challenge the underlying basis.
The practical response for every U.S. taxpayer is proactive: accurate and complete reporting of all income reconciled against third-party data, robust contemporaneous documentation of all material transactions, prompt voluntary disclosure through the appropriate program where past non-compliance exists, and immediate engagement of an experienced U.S. tax lawyer at the first sign of IRS interest. In an AI-driven enforcement environment with a diminished human oversight layer, preparation and proactive compliance are the most effective tools available.Disclaimer: This article provides broad information. It is only accurate as of the posting date. It has not been updated and may be out-of-date. It does not give legal advice and should not be relied on as tax advice. Every tax scenario is unique to its circumstances and will differ from the instances described in the article. If you have specific legal questions, you should seek the advice of a U.S. tax lawyer.
Disclaimer: This article provides broad information. It is only accurate as of the posting date. It has not been updated and may be out-of-date. It does not give legal advice and should not be relied on as tax advice. Every tax scenario is unique to its circumstances and will differ from the instances described in the article. If you have specific legal questions, you should seek the advice of a U.S. tax lawyer.