US Anti-Discrimination Law Applied to AI (Title VII & ADA)
Rather than a new AI statute, US federal anti-discrimination law (Title VII, ADA) already reaches biased AI hiring and employment tools. It is the enforcement baseline for algorithmic discrimination in US workplaces.
Overview
Liability does not depend on any AI-specific rule; it runs through the ordinary disparate-impact framework, under which a claimant identifies a particular employment practice that causes a disparity and the employer must then show the practice is job related for the position and consistent with business necessity. Validation and recordkeeping standards come from the 1978 Uniform Guidelines on Employee Selection Procedures, which the federal enforcement agencies apply to any selection device, including the four-fifths rule as a rough screen for adverse impact. Disability claims run on a separate track that needs no group statistics at all: a qualification standard or test that screens out an individual with a disability, an assessment that measures an impairment rather than the skill it purports to measure, and a refusal to provide an alternative format each stand alone. Exposure arrives through a charge rather than an audit, so the practical trigger is an EEOC charge or, after a right-to-sue notice, a private action. Federal posture shifted in 2025: the EEOC withdrew its 2022 ADA and 2023 Title VII technical assistance on algorithmic selection, and Executive Order 14281 directed agencies to deprioritise disparate-impact enforcement, but neither step repealed Title VII, the ADA or 29 CFR Part 1607, and private plaintiffs continue to litigate, with Mobley v. Workday allowing claims to proceed against a screening vendor on the theory that it acted as the employer's agent.
Key dates
- Jul 2, 1965Title VII enforcement begins
- May 18, 2023EEOC technical assistance on algorithmic selection issued (withdrawn January 2025)
- Jan 27, 2025EEOC removed its AI technical assistance from eeoc.gov following Executive Order 14179
Risk areas addressed
Who it applies to
Title VII and the ADA reach employers with 15 or more employees, along with employment agencies and labour organisations, while the ADEA sets its threshold at 20 or more; federal contractors carry parallel obligations through their contract terms. Screening vendors, assessment providers and applicant tracking suppliers can be pulled in as agents of the employer rather than treated as arm's length third parties, which is the theory currently proceeding in Mobley v. Workday in the Northern District of California. Employers below the headcount thresholds sit outside these statutes, as do certain religious organisations and Indian tribes, though state and city laws frequently apply at lower thresholds and Title VII can reach US citizens employed abroad by American-controlled employers.
Key obligations
- Do not use a selection procedure that has an adverse impact on the hiring or promotion opportunities of any race, sex or ethnic group unless the procedure has been validated in accordance with the Uniform Guidelines (29 CFR § 1607.3(A)).
- Where two procedures are substantially equally valid, use the one with the lesser adverse impact, and investigate suitable alternative procedures and methods of use as part of any validity study (29 CFR § 1607.3(B)).
- Treat a selection rate for any race, sex or ethnic group that falls below four-fifths of the rate for the highest-scoring group as evidence of adverse impact calling for validation or a change of procedure (29 CFR § 1607.4(D)).
- Maintain, for each job, records disclosing the impact your tests and selection procedures have by identifiable race, sex and ethnic group, together with the evidence of validity relied on wherever adverse impact is found (29 CFR §§ 1607.4(A), 1607.15(A)).
- Do not adjust scores, apply different cut-off scores, or otherwise alter the results of employment-related tests on the basis of race, colour, religion, sex or national origin (42 U.S.C. § 2000e-2(l)).
- Do not use qualification standards, employment tests or other selection criteria that screen out or tend to screen out individuals with disabilities unless the standard is job related and consistent with business necessity (42 U.S.C. § 12112(b)(6)).
- Select and administer assessments so results reflect the skill or aptitude being measured rather than an applicant's impaired sensory, manual or speaking skills, and provide reasonable accommodations including alternative assessment formats (42 U.S.C. §§ 12112(b)(5), 12112(b)(7); 29 CFR § 1630.11).
- Preserve personnel and employment records for one year from the record's creation or the personnel action, and preserve all records relevant to a charge until final disposition once a charge has been filed (29 CFR § 1602.14).
How to prepare
- Map each algorithmic step in the selection funnel, covering CV screeners, ranking and matching engines, video or game-based assessments, chatbots and interview scoring, and identify which step actually narrows the pool, because disparate impact attaches to a particular practice rather than to the process as a whole.
- Run an adverse-impact analysis on the prior 12 months of applicant flow data at each of those steps, calculating selection rates and impact ratios by race, sex and ethnicity yourself rather than accepting a vendor assurance.
- For any step showing adverse impact, assemble the validity evidence the Uniform Guidelines require, including a current job analysis and criterion or content validation, or swap the step for a substantially equally valid alternative with less impact.
- Build the disability route in separately: publish how to request an alternative assessment or accommodation before an assessment starts, train recruiters to grant it without penalising the applicant, and confirm the tool is not scoring reaction time, speech patterns, facial movement or typing where those are not genuine job requirements.
- Secure contractual rights from vendors to impact data, validation studies, audit results and cooperation in litigation, plus indemnities, on the basis that a vendor performing screening on your behalf may be treated as your agent.
- Retain applicant flow data, scores, model documentation and validation evidence for at least a year, and apply a litigation hold across all of it as soon as a charge or demand letter arrives.
Penalties & enforcement
Standard federal anti-discrimination remedies (back pay, damages, injunctive relief).
Enforced by: Equal Employment Opportunity Commission (EEOC)
US EEOC / Title VII (AI) is enforced by Equal Employment Opportunity Commission (EEOC). Its obligations are already live, so a gap is a present exposure rather than a future one. In practice that means knowing which of your AI systems fall in scope in employment & HR, holding assessments that speak to bias & discrimination, automated decisions and fundamental rights, and being able to produce that evidence on request.