IRS's Rapid AI Scale-Up for Audit Selection — GAO Finds a Twelvefold Jump in Use Cases and Serious Governance Gaps

The IRS uses AI-assisted models to analyze filed tax returns, assess them for noncompliance risk, and recommend cases for audit or further investigation, targeting returns from large corporations, complex partnerships, high-net-worth individuals, and users of digital assets.

The agency has been rapidly expanding its use of AI across audit selection, fraud detection and broader compliance operations.

GAO reviewed the IRS's AI use-case inventory, expected benefits documentation, and staffing capacity, publishing its findings as GAO-26-107522 ('Artificial Intelligence: IRS Actions Needed to Address Skills Gaps, Information Quality, and Strategic Management'), released March 24, 2026.

The number of active AI use cases at the IRS grew from just 10 in August 2022 to 126 by June 2025 -- a roughly twelvefold increase in under three years, with 61% of those use cases still in development as of the review. GAO found that the IRS's own inventory of its AI applications was incomplete, did not capture all the ways AI was being used, and did not clearly document how each tool was expected to benefit the agency. The report ties these governance gaps to a sharp contraction in the IRS workforce, which fell from more than 100,000 employees at the end of 2024 to approximately 75,000 by mid-2025, a reduction of roughly 25,000 positions.

GAO concludes the agency does not have enough skilled staff to adequately support, develop, or oversee its expanding portfolio of AI tools, and recommends the IRS strengthen strategic management of its AI programme alongside improving the completeness and quality of its AI-use inventory. The case illustrates a pattern also seen in the U.K. National Audit Office's July 2025 cross-government review of fraud and error analytics: public bodies can scale AI and data-analytics tools quickly, but proving and governing their actual return often lags well behind deployment.