This article summarises publicly available guidance from regulators and official sources. It is general educational information only and does not constitute legal or professional advice. Requirements vary by jurisdiction. Consult your regional authority or a qualified professional for advice specific to your situation.
An AI hiring tool doesn't need to be built with any intent to discriminate to create real Title VII exposure, it just needs to produce results that disproportionately screen out a protected group, even from a facially neutral process. The EEOC's guidance on this is specific: your business remains liable for that outcome even if you relied entirely on your vendor's assurance the tool was tested and clean.
In short: The EEOC's guidance on AI in employment selection focuses on adverse impact, whether a facially neutral AI tool disproportionately screens out people based on a Title VII-protected characteristic. Employers are expected to evaluate this themselves, not just rely on vendor claims, and the traditional four-fifths rule is a useful, though not definitive, way to check. If the vendor's bias testing turns out to be wrong, the employer can still be held liable.
The plain-English answer
Using an AI hiring tool is legal, and most tools aren't deliberately built to discriminate. The risk the EEOC's guidance addresses is a facially neutral tool that, in practice, disproportionately screens out candidates who share a protected characteristic, race, sex, age, and others covered by Title VII, regardless of intent. Your business's obligation isn't to prove the tool was built with good intentions, it's to check the actual outcomes it produces against your real applicant pool.
Why this matters for your business
A 25-person logistics company using an AI ranking tool for every applicant is a genuinely common setup, and relying on the vendor's assurance that the tool was "tested for bias" is the natural, reasonable-seeming default, most businesses don't have the in-house expertise to run their own statistical bias testing from scratch. The EEOC's guidance makes clear this default isn't sufficient on its own: the employer can still be held liable for disparate impact discrimination even where the vendor represented the tool as adequately tested, if that representation turns out to be wrong.
This isn't a reason to distrust every AI hiring vendor, most operate in good faith. It's a reason to build a real, independent check into your own hiring process rather than treating a vendor's marketing claim as the end of the compliance question.
What the EEOC actually says
The EEOC's guidance, "Assessing Adverse Impact in Software, Algorithms, and Artificial Intelligence Used in Employment Selection Procedures Under Title VII," focuses specifically on unintentional, facially neutral discrimination, disparate impact, rather than deliberate bias. It recommends employers conduct a thorough and ongoing evaluation of whether their AI selection tool disproportionately excludes people based on a protected characteristic, and it's explicit that vendor due diligence, asking whether the vendor has evaluated the tool for disparate impact, doesn't shift the ultimate liability away from the employer if the vendor's assessment turns out to be incorrect.
The guidance points to the traditional four-fifths rule as a useful, though not definitive, way to check: if the selection rate for a protected group is less than four-fifths (80%) of the rate for the highest-selected group, that's generally treated as evidence of adverse impact, though an employer can still defend a tool that produces this pattern if it can show the tool is job-related and consistent with business necessity. The guidance also recommends regular, ongoing validation and testing, not a one-time check at rollout.
What this looks like in practice
Picture the talent acquisition manager at the logistics company, having rolled out the AI ranking tool on the vendor's assurance that it was tested for bias, with no independent check of the business's own applicant data since launch. That's a real gap under the EEOC's guidance, the vendor's claim, however genuine, doesn't satisfy the employer's own obligation to evaluate actual outcomes.
Running the four-fifths check against the company's own applicant data, comparing the AI tool's selection rate for each protected group against the highest-selected group, is a manageable exercise even without a statistician on staff, most applicant tracking systems can export the underlying data. If a group's selection rate comes in under 80% of the top group's, that's the signal to investigate further, either the tool needs adjustment or the business needs to be ready to show the selection criteria are genuinely job-related.
What you can do about it
A practical process for checking your own AI hiring tool's real-world impact:
- Ask your AI vendor specifically what bias testing they've done, and request the methodology and results, not just a general assurance.
- Run your own four-fifths check against your actual applicant data periodically, not just once at rollout.
- If a selection rate gap shows up, investigate before it becomes a pattern, don't wait for a complaint to prompt the review.
- Keep documentation of both the vendor's testing and your own checks, this is the evidence that matters if the tool's outcomes are ever questioned.
- Be ready to show your selection criteria are job-related and consistent with business necessity if a gap does show up and you intend to keep using the tool as-is.
If your hiring process could evaluate any candidate who resides in New York City, a separate, specific law applies on top of this Title VII guidance. See our guide on NYC Local Law 144's AI hiring bias audit requirement.
Methodology (Real-World, Verified)
We test AI tools against real SMB workflows: the tasks a 20-person business actually uses AI for, not enterprise demos. Pricing is verified at the vendor's published rates, with local-currency conversions noted where relevant. Compliance notes reference the legislation and regulatory guidance relevant to each article's region. Every tool is judged on one question: could a business with no dedicated IT department actually pick this up and use it on Monday morning.
Related reading: our AI governance by region.
Try our free AI Compliance Checker to check whether your AI tools meet your compliance obligations.
Does the EEOC's guidance require us to hire a statistician to test our AI hiring tool?
Not necessarily, the four-fifths rule is a relatively accessible check most businesses can run against their own applicant data using a spreadsheet. More sophisticated statistical testing is available if a real gap shows up and needs deeper investigation, but it's not a prerequisite to doing a basic first check.
What if our AI vendor refuses to share their bias testing methodology?
Treat that as a real concern. The EEOC's guidance expects employers to actually evaluate this, and a vendor unwilling to share how they tested for bias makes that evaluation much harder to do properly, worth weighing seriously against the tool's other benefits.
Does passing the four-fifths test mean our AI hiring tool is fully compliant?
It's a useful signal, not a guarantee. The EEOC's guidance describes it as helpful but not definitive, other forms of evidence can still show adverse impact even where the four-fifths threshold isn't crossed. Treat it as a first check, not the final word.
Does this guidance cover intentional discrimination too?
No, the EEOC's AI-specific guidance focuses on unintentional disparate impact from facially neutral tools. Intentional discrimination through AI, or any other means, is covered by existing Title VII disparate treatment law separately, and remains just as prohibited.
The information in this article is general in nature. It reflects a summary of publicly available guidance and does not constitute legal, privacy, or professional advice. Your obligations will depend on your specific situation, jurisdiction, and business circumstances. Do not rely on this article as a substitute for qualified legal or professional advice.
Could any of your candidates be NYC residents, regardless of where your business is based?
Check Local Law 144