IRS audit selection algorithms disproportionately target Black taxpayers

Stanford researchers found that Black taxpayers were audited at 2.9 to 4.7 times the rate of non-Black taxpayers, with the disparity most pronounced among EITC claimants. The IRS confirmed these findings in a May 2023 letter to Congress after an internal review, and multiple outlets corroborated the disparity and its attribution to audit-selection algorithms.

United States Internal Revenue Service (IRS) · Incident Jan 31, 2023 · Indexed Jun 5, 2026 · 4 sources

The racial disparity in audit selection is driven by algorithms that prioritize the likelihood of underreporting over the magnitude of the underreported income.
What
Stanford researchers found that Black taxpayers were audited at 2.9 to 4.7 times the rate of non-Black taxpayers, with the disparity most pronounced among EITC claimants.
Incident date
Jan 31, 2023
Who
United States Internal Revenue Service (IRS)
Failure mode
Identity & Access Drift
AI surface
Agentic Workflow
Severity
High

What happened

A Stanford University research team found that Black taxpayers receive IRS audit notices 2.9 to 4.7 times more often than non-Black taxpayers. The IRS confirmed these findings in a May 2023 letter to Congress after an internal review of its auditing process. The bias was found to be most extreme among claimants of the Earned Income Tax Credit.

What broke inside the model

Failure path · mode profile · Identity & Access Drift
  1. 01 · TriggerAn agent operates with granted credentials.
  2. 02 · Model stepIt reaches for scope it was never assigned.
  3. 03 · Control gapNo runtime check binds it to its role.
  4. 04 · FailureThe agent acts outside its authority.
  5. 05 · ConsequencePrivileged actions run with no oversight.

The agent's actions drift outside the scope it was granted.

The audit selection algorithms focused on the probability of an error occurring rather than the monetary magnitude of the underreported tax. This approach disproportionately flags low-income taxpayers, who are more likely to be Black, for audits.

Public visibilityHigh
Regulatory exposureActive
Customer impactMany customers
Financial impactUnknown
Time to disclosureMonths
  1. PressIRS Disproportionately Audits Black Taxpayerslaw.stanford.edu
  2. PressBlack Americans Are Much More Likely to Face Tax Auditsnytimes.com
  3. PrimaryCommissioner Werfel Letter on Audit Selectionirs.gov
  4. PressIRS Confirms Stanford Study of Racial Bias in Auditslaw.stanford.edu
Permalinkhttps://failureindex.ai/failures/irs-audit-selection-algorithms-disproportionately-targe
CitationAI Failure Index. "IRS audit selection algorithms disproportionately target Black taxpayers" (FI-0249). Realm Labs. https://failureindex.ai/failures/irs-audit-selection-algorithms-disproportionately-targe (indexed Jun 5, 2026).
Share cardA branded image of this record for posts and slides.

Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0249. Full dataset at /data.

Note from Realm Labs, the Index steward

How Realm would have caught this

Controls for this failure mode
  • OmniGuard
  • AgentRealm

Realm can bind an agent's actions to the identity and scope it was assigned and flag the moment it reaches for access beyond its task, so inherited or discovered permissions do not quietly become a destructive action.