What US employers are actually liable for when AI makes hiring decisions.

Your AI screener rejected someone illegally and you probably don't know it.

The EEOC’s AI and hiring guidance: what US employers are legally exposed to right now

The federal guidance is gone. The legal exposure isn’t.

In January 2025, after Trump’s executive order directing agencies to clear Biden-era AI rules that conflicted with an “AI dominance” agenda, the EEOC pulled its May 2023 technical assistance document on AI compliance under Title VII from its website, along with other AI-related publications. Some employers read that as a green light. It wasn’t.

Title VII, the ADA, the ADEA, and every other federal anti-discrimination statute the guidance was interpreting are still fully in force. Removing the guidance didn’t change the underlying law. It just removed the agency’s own explanation of how it planned to apply that law to AI tools. For employers, that’s arguably a worse position to be in, not a better one.

Here’s what the exposure picture actually looks like.

The underlying legal theory hasn’t moved

The 2023 EEOC guidance didn’t invent new obligations. It applied a 1979 framework (the Uniform Guidelines on Employee Selection Procedures) to AI-driven hiring tools and reached a conclusion that was always going to be uncomfortable: the EEOC treats employer use of algorithmic decision-making tools as an employment “selection procedure” under Title VII, meaning any tool used to make or inform decisions about hiring, promotion, or termination is subject to the same disparate impact analysis as a written test or structured interview.

Under the four-fifths rule, adverse impact is generally indicated where the selection rate for any protected group is less than 80% of the rate for the highest-selecting group. That threshold applies whether the selection tool is a multiple-choice aptitude test from 1985 or a machine learning resume screener deployed last quarter.

The part that catches most founders off guard: an employer can be held responsible under Title VII for selection procedures that use an algorithmic tool even if the tool was designed or administered by another entity, such as a software vendor. Buying an off-the-shelf AI hiring product and letting the vendor run it doesn’t transfer the liability. You’re still the employer. The obligation to check whether the tool produces discriminatory outcomes sits with you.

The EEOC encouraged employers to conduct self-analysis on an “ongoing basis” and make proactive changes if appropriate, not just a one-time pre-deployment check. That framing matters. A tool that passes a bias audit in year one can drift as the model is retrained or as your applicant pool changes.

What enforcement looks like in practice

In August 2023, the EEOC settled its first AI hiring discrimination case. In EEOC v. iTutorGroup, it alleged that the company’s hiring software automatically rejected older job applicants in violation of the Age Discrimination in Employment Act. The EEOC’s claims were that iTutorGroup programmed its software to reject female job applicants aged 55 or older and male applicants aged 60 or older. The settlement required iTutorGroup to pay $365,000 to over 200 applicants whose applications were rejected because of their age.

A $365,000 settlement isn’t a catastrophic number for a large enterprise. For a growth-stage company or a staffing agency operating on thin margins, it’s a different conversation. And that case involved age discrimination that was essentially hardcoded into the system. The more common risk is subtler: a resume screener that was trained on historical hiring data ends up de-prioritising candidates from certain universities, zip codes, or career trajectories that correlate with race or gender, and nobody notices until a pattern of rejections accumulates.

---

AI Governance in HR and Hiring - Spring 2026

---

The current EEOC posture

Under Chair Lucas, the EEOC has directed the agency to close investigations that rely solely on disparate impact theory. That’s a real shift. The prior administration treated disparate impact as a primary enforcement hook for AI discrimination claims. The current agency is less interested in it.

A private right of action still exists under Title VII, and state and local authorities may still act on disparate impact claims. So while federal enforcement appetite for AI-related disparate impact cases has cooled, the legal mechanism for a plaintiff to sue, or for a state AG to investigate, remains intact. Employers who stopped running adverse impact analyses because the EEOC went quiet are taking a risk they may not have properly priced.

Where state law fills the gap