The IRS Is Testing AI Tools to Decide Who Gets Audited

Why It Matters

AI‑driven audit selection could dramatically increase the IRS’s collection efficiency while raising questions about fairness and data privacy, reshaping compliance dynamics for millions of taxpayers.

Key Takeaways

• IRS contracts AI, spending ~ $20 million annually.
• 126 AI use cases; two‑thirds launched 2022‑2025.
• Risk‑score models flag returns before human audit decision.
• Audit openings fell to 205k in FY2025, down from 335k.
• Critics warn of bias and privacy risks in AI selection.

Pulse Analysis

The IRS’s pivot to artificial‑intelligence reflects a broader governmental push to compensate for staffing shortfalls with technology. After the Trump administration slashed audit personnel by roughly one‑third, the agency’s enforcement revenue slipped about 5 percent in the last fiscal year. In response, the service has accelerated AI adoption, logging 126 distinct use cases—most of them launched after the Inflation Reduction Act boosted funding. Current contracts with vendors such as Microsoft, Palantir, and IBM fund an estimated $20 million a year in AI development, positioning the IRS to analyze millions of returns simultaneously.

At the heart of the new system are risk‑scoring algorithms trained on historic audit data. Each return receives a probability score indicating potential non‑compliance, which then triggers a human review before any audit is issued. This hybrid model promises to concentrate limited auditor resources on the most financially significant cases, potentially reversing the decline from 335,000 audits in FY2024 to 205,000 in FY2025. However, the reliance on algorithmic decision‑making introduces concerns about inadvertent bias, especially if training data reflect past enforcement disparities. Privacy advocates also warn that expanded data mining could expose sensitive taxpayer information.

For the broader tax ecosystem, AI‑enhanced enforcement signals a shift toward more data‑driven compliance strategies. Tax professionals may need to adjust filing practices to mitigate algorithmic red flags, while legislators will likely scrutinize the transparency and accountability of these tools. If the IRS can demonstrate that AI improves collection without compromising fairness, it could set a precedent for other agencies seeking efficiency gains. Conversely, any missteps could spark regulatory pushback and erode public trust in the tax system.