IRS Leverages AI for Smarter Tax Audits

IRS Uses AI to Rank Tax Returns for Audits

The Internal Revenue Service (IRS) is increasingly using artificial intelligence (AI) to help identify tax returns that may warrant an audit. This technology aims to improve efficiency and focus compliance efforts on areas where they are most needed. While AI plays a significant role in ranking and prioritizing these returns, human review remains a critical step before any final audit decision is made. This approach is part of a broader effort to modernize tax collection and enforcement.

The IRS has categorized its AI systems that influence audit selection or scope as “high-impact.” This designation means these systems are subject to strict governance policies, ensuring accountability and oversight. The agency is not handing over audit decisions entirely to machines; instead, AI acts as a sophisticated tool to analyze vast amounts of data and flag potential issues. This allows tax officials to concentrate their resources more effectively.

How AI Assists in Audit Selection

AI systems are being integrated into the IRS’s compliance work to enhance the process of selecting tax returns for review. These tools build upon older analytical methods, adding speed, scale, and advanced pattern detection. The goal is to identify complex issues, potential errors, or instances of fraud more quickly and accurately than traditional methods alone might allow.

The IRS has been exploring various applications of AI in this area. This includes developing models to examine complex partnerships and high-risk returns. One contract involved testing three AI models for case selection, demonstrating a commitment to experimenting with and refining these technologies. These efforts aim to make the audit selection process more dynamic and responsive to current financial landscapes.

Traditional Audit Triggers Remain Relevant

Despite the advancements in AI, many of the factors that have historically drawn IRS scrutiny are still important. These include returns with high incomes, complex financial structures, aggressive claims for credits or deductions, and unusual charitable contributions. AI appears to work by sorting and prioritizing these familiar audit triggers across a large volume of tax filings.

For example, a complex return with unusual deductions would likely have attracted attention in the past. AI can now help determine how quickly such a return rises to the top of the review pile and which specific parts of the filing should receive closer examination. This integration of AI does not replace traditional audit markers but rather enhances the speed and precision with which they are assessed.

Addressing Concerns About Algorithmic Bias

As AI becomes more integrated into sensitive government functions like tax enforcement, concerns about potential bias are being raised. Studies have shown that certain demographic groups have historically been audited at higher rates. Independent analyses have identified the possibility of “unintentional algorithmic biases” within automated systems.

The IRS acknowledges these concerns and is working within its AI governance policy to manage them. While the agency states it does not engage in deliberate discrimination, the potential for bias in AI systems is a recognized challenge. Lawmakers and oversight bodies are likely to continue pressing for clear rules regarding the testing, documentation, and review of these AI systems to ensure fairness and accountability in the audit selection process. The “high-impact” classification of these systems is intended to address some of these oversight needs.

The Role of Human Review in the Process

A crucial aspect of the IRS’s approach to AI in audit selection is the continued involvement of human reviewers. AI systems are designed to inform or influence audit decisions, but they do not make the final call. A human tax professional must review the flagged return and the AI’s assessment before any audit is initiated.

This human oversight is intended to ensure accountability and to provide a layer of judgment that AI alone cannot replicate. It acts as a safeguard against fully automated decisions, ensuring that a person is ultimately responsible for deciding whether a taxpayer will be audited. However, the recommendation or ranking provided by the AI system can still significantly shape the reviewer’s focus and the direction of the subsequent examination.

Future Directions for AI in Tax Administration

The IRS is clearly moving towards a deeper integration of AI within its compliance and enforcement operations. The agency’s efforts range from broad development of AI use cases to specific, targeted contracts for AI tools. This direction suggests a future where data analytics and improved technology will play an even larger role in focusing compliance efforts.

The ongoing debate centers on balancing the efficiency gains offered by AI with the imperative of fairness and accountability. Policymakers face the challenge of enabling the broader use of AI in tax administration while ensuring that audit decisions remain subject to human judgment and that potential biases are effectively mitigated. The IRS’s current AI governance policy, which labels audit-related systems as “high-impact,” is a step in this direction, but the effectiveness of safeguards against bias remains a key question for the future.

Frequently Asked Questions