In 2023, a resume-screening algorithm at iTutorGroup was found to have been programmed to reject female applicants over 55 and male applicants over 60. The EEOC brought suit; iTutorGroup settled for $365,000. The specific facts of that case are unusual — most employer-side AI discrimination is not this deliberate — but the legal principle the case established is broadly applicable: Title VII of the Civil Rights Act applies to AI-driven employment decisions the same way it applies to human ones.
For a small business using AI hiring tools — resume parsers, candidate ranking systems, video-interview AI, personality assessments, or any algorithm that influences who gets an interview or a job — the employer is legally responsible for the tool's decisions. “The algorithm decided” is not a defense to disparate-impact discrimination or disparate-treatment discrimination under Title VII, the ADEA (age discrimination), the ADA (disability), or state equivalents.
Most small businesses using these tools don't understand the compliance framework this creates. The result: exposure to lawsuits, regulator actions, and settlement demands that most business owners never see coming until they arrive.
This is a practitioner's guide to what small business owners need to know about AI hiring tools and employment discrimination law.
The core legal principle
Title VII prohibits employment discrimination based on race, color, religion, sex, or national origin. The ADEA prohibits age discrimination against workers 40 and older. The ADA prohibits disability discrimination.
Two doctrines matter for AI tool use:
Disparate treatment — intentional discrimination against a protected class. AI tools can violate this if programmed with discriminatory criteria (the iTutorGroup case) or if their outputs are used to discriminate.
Disparate impact — facially neutral policies or practices that fall more heavily on a protected class without justification. This is the more common AI-related issue: a tool trained on historical hiring data may replicate historical discrimination patterns even without explicit discriminatory intent.
The EEOC's 2023 and 2024 guidance clarified that both doctrines apply to AI-driven employment decisions. Enforcement in 2025 and 2026 has been active.
The four-fifths rule applied to AI outputs
The EEOC's long-established “four-fifths rule” (also called the 80% rule) provides a screening standard for disparate impact. Under the rule:
If a hiring practice results in one protected class being selected at less than 80% of the rate of the highest-selected class, that's a prima facie showing of disparate impact.
Example. An AI screening tool selects 60% of male applicants for interviews but only 40% of female applicants. Female selection rate (40%) / male selection rate (60%) = 67% — less than 80%. Prima facie disparate impact.
The employer's defense is to show the tool is job-related and consistent with business necessity, and that no less-discriminatory alternative exists. That's a substantive defense requiring specific evidence — often including a formal “bias audit” of the tool's outputs.
The bias audit obligation
Some jurisdictions now require formal bias audits for AI hiring tools:
New York City Local Law 144 requires employers using “automated employment decision tools” (AEDTs) in NYC to conduct annual independent bias audits, disclose the audit results, and notify candidates that AI is being used. The rule has specific requirements for who can conduct the audit and what it must measure.
Illinois Artificial Intelligence Video Interview Act requires employer disclosure when AI is used to analyze video interviews, plus consent from candidates.
California and other states have proposed similar frameworks; the Colorado AI Act (effective 2027) will impose broader documentation requirements on developers and deployers of “high-risk” AI systems including employment tools.
Even in jurisdictions without specific AI-hiring statutes, the general Title VII framework requires the employer to defend the tool's outputs. Documentation of bias auditing, whether or not statutorily required, is often part of that defense.
The vendor question
Most small businesses don't build AI hiring tools — they buy or subscribe. But buying doesn't transfer legal responsibility. The employer using the tool is generally the party sued or investigated, regardless of who built the tool.
This makes vendor selection a compliance decision. Questions to ask any AI hiring tool vendor before adopting:
On the tool itself:
- What specific data was the tool trained on? Are there known bias risks in the training data?
- Has the tool been subjected to independent bias auditing? What did the audits find?
- What documentation can the vendor provide about the tool's outputs by protected class?
On contractual protection:
- Does the vendor indemnify the employer against discrimination claims arising from tool use?
- Does the vendor provide the audit and outcome data needed to defend Title VII challenges?
- What's the vendor's obligation to update or improve the tool if bias is identified?
Colorado's AI framework (effective 2027) makes some of these vendor obligations statutory — developers of “high-risk” AI systems will be required to provide deployers with documentation supporting compliance. Other states are likely to follow. For a broader treatment of AI vendor contract negotiation, see our AI vendor contract checklist.
The specific AI hiring tools that create the most risk
Not all AI hiring tools carry equal risk. Some patterns generate more exposure than others:
Highest-risk uses:
- Resume screening algorithms that reject candidates without human review
- Video interview AI that scores candidates on communication, personality, or other subjective factors
- Personality or “culture fit” assessments used as hiring gates
- Algorithmic ranking that determines who advances in the process
Moderate-risk uses:
- Sourcing tools that identify candidates from public profiles (subject to source-population bias)
- Job description optimization tools that may inadvertently signal preferred demographics
- Scheduling tools that make assumptions about candidate availability
Lower-risk uses:
- Administrative tools (interview scheduling, offer letter generation)
- Analytics tools that report on hiring patterns without making decisions
- Onboarding automation for already-hired employees
Small businesses should scrutinize higher-risk uses most carefully, and consider whether the productivity gain justifies the compliance overhead.
The practical framework for small business
Six steps for small businesses using or considering AI hiring tools:
- Inventory current use. What AI or algorithmic tools are involved in the hiring process, from job posting through offer? Include tools embedded in ATS platforms that may not be obviously “AI.”
- Identify the decision points. Where does the tool make or influence decisions about candidates? These are the compliance-critical points.
- Require human review at decision points. AI tools that recommend or rank should be reviewed by a human before rejection or advancement. Human review doesn't eliminate liability but establishes that decisions weren't made purely algorithmically.
- Document reasoning. For any adverse decision, document the specific reason. If the AI tool contributed, note what it flagged and how the human decision-maker weighed it.
- Monitor outcomes by protected class. Track hiring outcomes by race, gender, age, and other protected characteristics. If your outcomes suggest disparate impact, understand why before someone else identifies the pattern.
- Verify vendor compliance posture. Understand what your vendors have done on bias auditing and what documentation they can provide. Get contractual protection where possible.
When to consult employment counsel
Situations warranting attorney involvement:
- Adopting any new AI hiring tool
- Post-hoc discovery of hiring patterns suggesting disparate impact
- EEOC charge, state agency complaint, or plaintiff's demand letter
- Operations in NYC, Illinois, or other jurisdictions with specific AI hiring rules
- Multi-state operations where compliance framework varies
An employment attorney with AI-tool experience is not always easy to find at the small-business fee level. The Silverton view: the fee for proactive compliance guidance is trivial compared to the cost of defending a class-action disparate-impact case.
Frequently Asked Questions
If my AI hiring tool discriminates, am I responsible or is the vendor?
You (the employer) are generally responsible for the tool's outputs under Title VII and related laws. Vendor contracts may provide indemnity, but the plaintiff or the EEOC will typically pursue the employer. Buying an AI tool doesn't transfer legal responsibility for hiring decisions.
Does the EEOC's four-fifths rule apply to AI tools?
Yes. The EEOC has clarified that the four-fifths rule (screening for disparate impact when a protected class is selected at less than 80% of the rate of the highest-selected class) applies to AI-driven hiring decisions the same way it applies to human ones. AI outputs that fail the four-fifths test create prima facie disparate impact.
Do I have to tell candidates I'm using AI in hiring?
Depends on jurisdiction. New York City's Local Law 144 requires disclosure of automated employment decision tools. Illinois requires disclosure and consent for AI video interview analysis. California and other states have proposed similar rules. Even where not statutorily required, disclosure is often a defensive practice — it eliminates one argument in a subsequent challenge.
Do bias audits actually protect me legally?
They help but don't eliminate risk. A well-conducted bias audit showing the tool doesn't produce disparate impact is evidence in the employer's defense. An audit showing disparate impact — that the employer didn't address — is evidence for a plaintiff. Whether to conduct audits, use audit-vendor-certified tools, and document findings should be strategic decisions made with counsel.
What should I do if I discover my AI hiring tool has been producing discriminatory outcomes?
Consult employment counsel immediately. Depending on the situation, appropriate responses may include: pausing use of the tool, correcting the specific pattern, notifying affected candidates, addressing exposure with your vendor, and updating hiring processes going forward. Trying to handle this without counsel typically compounds the problem.
This article draws from the forthcoming Silverton Publishing book AI in Business and Law: A Practical Guide to Using Artificial Intelligence Without Getting Sued. It is general information only, not legal advice. Employment discrimination law is complex and fact-specific. Consult a qualified employment attorney for guidance on your specific situation. For ongoing updates on AI employment law developments after this article's publication, see silvertonpublishing.com/ai-current.
This article is for educational purposes only and does not constitute legal, tax, or financial advice. Consult a qualified professional for guidance specific to your situation.