AI Enters Audit Selection
A recent government report documents the IRS's expanding use of artificial intelligence in selecting tax returns for audit. The agency is applying AI to identify returns with higher likelihoods of non-compliance, targeting a more efficient path toward closing the tax gap.
The Consistency and Transparency Standard
The report emphasizes that AI use in this context must satisfy two operational standards:
- Consistency — models and selection logic must apply uniformly across taxpayers so enforcement remains equitable
- Transparency — the agency must be able to explain how AI contributes to audit decisions in a way that sustains public trust and supports taxpayer rights
Meeting both standards simultaneously is genuinely difficult with modern ML systems, which is why the report treats them as central concerns rather than procedural formalities.
What the IRS Is Trying to Achieve
The stated goal is to:
- Streamline audit selection through higher-signal targeting
- Concentrate limited enforcement resources on the cases most likely to produce adjustments
- Close the tax gap more efficiently without broadly expanding audit volume
Safeguards Matter
For AI-assisted audit selection to function well, appropriate safeguards are essential — ensuring taxpayer rights remain protected, that selection is not arbitrary, and that model outputs are supported by human review before enforcement actions occur.
Implications
For tax practitioners, the shift toward AI-assisted selection means that audit triggers will be less intuitive and potentially harder to anticipate. Maintaining strong documentation, consistent reporting positions, and clean compliance records remains the best protection — regardless of how the selection model evolves.