# Disparate impact from AI-skill requirements[^about] How AI-skill requirements can create disparate impact risk in hiring and promotions under Title VII, the ADEA, and national-origin laws. ## How much AI competence can employers require before hiring or promotion? {#how-much-ai-competence-baseline} **Short answer.** It depends, because AI literacy can be part of a role, but a pre-hire AI fluency gate still has to be tied to the job and checked for disparate impact. The governing law is older than the tools. Title VII forbids employers to `limit, segregate, or classify` applicants or employees in ways that deprive them of employment opportunities because of race, color, religion, sex, or national origin, and Section 703(k) codifies the familiar burden-shifting rule for disparate-impact claims.limit, segregate, or classify [^42-u-s-c-section-2000e-2-k] The classic cases still frame the analysis: *Griggs v. Duke Power Co.*, 401 U.S. 424 (1971), *Albemarle Paper Co. v. Moody*, 422 U.S. 405 (1975), *Watson v. Fort Worth Bank & Trust*, 487 U.S. 977 (1988), and *Connecticut v. Teal*, 457 U.S. 440 (1982). None involved LLMs. That is partly why they still matter. They treat screening tools functionally, not by software label. [^griggs-v-duke-power-co-401-u-s-424-1971][^albemarle-paper-co-v-moody-422-u-s-405-1975][^watson-v-fort-worth-bank-trust-487-u-s-977-1988][^connecticut-v-teal-457-u-s-440-1982] The Uniform Guidelines on Employee Selection Procedures make the mapping to AI-skill gates fairly direct. The EEOC's 1979 Q&A says the Guidelines apply to "all selection procedures used to make employment decisions"[^eeoc-questions-and-answers-to-clarify-and-provid], including interviews, review of experience or education, work samples, job requirements, and evaluations of performance. [^eeoc-questions-and-answers-to-clarify-and-provid] So a filter framed as `AI proficiency required`, a prompt-writing exercise, or a screen for prior LLM usage is not outside the doctrine. It is inside the doctrine under a newer name. The operational screen is still the four-fifths rule, but not as a safe harbor. The Guidelines say a selection rate below four-fifths (4/5) (or eighty percent) of the highest group's rate will generally be regarded as evidence of adverse impact. The next sentence matters just as much: smaller differences can still count when they are statistically and practically significant, and larger differences can fail to matter when the numbers are too small. [^29-c-f-r-part-1607] That is why a company can clear the rough 0.80 ratio and still have a problem, or fail it and still argue the sample is too thin to mean much. The EEOC's current AI materials say the old laws apply to new tools. Its 2024 one-pager says federal employment-discrimination laws apply to AI just as they apply to other employment practices, including recruiting, screening, pay, promotion, and termination decisions. [^eeoc-what-is-the-eeoc-s-role-in-ai][^eeoc-employment-discrimination-and-ai-for-worker] There is still no reported federal appellate decision, at least in the reviewed sources, squarely about `AI skill required` as a hiring gate. The present issue is old doctrine applied to a new filter, not a fresh statutory regime. There is not yet a thick stack of Am Law commentary aimed specifically at `must already know AI` job postings. The firms are triangulating the issue from two adjacent areas: the EEOC's 2023 AI selection-procedure guidance, and the newer employer commentary on AI literacy and job design. That makes the consensus narrower than the headlines. It is not `AI requirements are illegal`. It is `AI requirements are selection criteria, so ordinary validation and disparate-impact rules attach`. [^littler-eeoc-issues-guidance-on-use-of-artificia][^mayer-brown-eeoc-issues-title-vii-guidance-on-em][^fisher-phillips-eeoc-s-latest-ai-guidance-sends][cite source=Paul Hastings, \AI\ in Employment Law: EEOC Issues Title VII Guidance url=https://www.paulhastings.com/insights/client-alerts/ai-in-employment-law-eeoc-issues-title-vii-guidance] Littler, Mayer Brown, Fisher Phillips, and Paul Hastings are closely aligned on the doctrinal core. They all describe AI-driven employment tools as selection procedures under Title VII and the Guidelines. Fisher Phillips says the four-fifths rule `can be applied to AI selections`. Paul Hastings calls disparate impact the `principal risk` when AI produces practically and statistically significant harm to protected groups. Mayer Brown and Littler say much the same thing in more measured language: AI does not create a third category between lawful and unlawful screening. It drops into the same bucket as other screening criteria. [^fisher-phillips-eeoc-s-latest-ai-guidance-sends][cite source=Paul Hastings, \AI\ in Employment Law: EEOC Issues Title VII Guidance url=https://www.paulhastings.com/insights/client-alerts/ai-in-employment-law-eeoc-issues-title-vii-guidance] [^mayer-brown-eeoc-issues-title-vii-guidance-on-em][^littler-eeoc-issues-guidance-on-use-of-artificia] Mintz is the closest thing in the source set to direct commentary on the skill-requirement version of the problem. Its April 2026 piece treats the Department of Labor's AI Literacy Framework as a signal that regulators see AI literacy becoming a baseline workforce expectation. But the same article goes in a narrower direction when it reaches actual role design: it distinguishes vague `AI proficiency` language from job descriptions that identify the specific tools, platforms, or competencies expected for the role. [^mintz-ai-in-the-workplace-issue-spotting-for-emp][^the-u-s-department-of-labor-s-artificial-intelli] That is not quite a disagreement with the 2023 AI-selection memos. It is more revealing than that. It suggests the employer-side bar is already treating `everyone must already be good at AI` as too blunt a description to carry much weight on its own. The first consequence is that `AI literacy` and `preexisting AI fluency` are not the same requirement. The Department of Labor's framework says "every worker will need baseline AI literacy skills"[^the-u-s-department-of-labor-s-artificial-intelli], but it also says many roles will require different depths of knowledge and that employers may need to define the specific AI skills appropriate for each role. [^the-u-s-department-of-labor-s-artificial-intelli] That tends to separate two propositions that are often collapsed in practice: `this job now includes AI-enabled work` and `only people who already use AI heavily should be allowed through the gate`. The practical split is therefore less `AI role` versus `non-AI role` than `specific capability` versus `proxy for prior opportunity`. A role built around supervising model outputs, maintaining AI-assisted workflows, or auditing AI systems may have a stronger job-relatedness story than a generalist role that suddenly demands `prompt engineering` because the company wants everyone to become AI-first. [^the-u-s-department-of-labor-s-artificial-intelli][^mintz-ai-in-the-workplace-issue-spotting-for-emp] The former can at least be tied to concrete tasks. The latter is more likely to look like a market-sentiment filter. The DOL framework says every worker will need baseline AI literacy, but it also says roles require different depths of knowledge. No court has yet mapped that distinction onto a pre-hire `must already be fluent` requirement. [^the-u-s-department-of-labor-s-artificial-intelli] ## What alternatives should employers consider before using an AI-fluency screen? {#alternatives-to-ai-fluency-screen} **Short answer.** Usually, employers should compare the AI screen against training after hire, work samples with tool access, and narrower tests of judgment over AI output. The alternative-practice point also matters more here than it first appears. EEOC testing guidance says that if a selection procedure screens out a protected group, the employer must determine whether there is an equally effective alternative with less adverse impact. [^eeoc-employment-tests-and-selection-procedures] For AI-skill gates, the obvious comparison is not theoretical. It is usually some combination of training after hire, a job-specific work sample with AI use permitted, or an evaluation of editing and judgment over AI output rather than prior exposure to a named tool. Existing law clearly allows plaintiffs to point to alternatives with less adverse impact. What is not yet settled is whether the strongest comparator for an AI-fluency screen is training after hire, a work sample with tool access allowed, or a narrower test of judgment over AI output. [^42-u-s-c-section-2000e-2-k-2][^eeoc-employment-tests-and-selection-procedures] ## How does the ADEA apply to AI-native skill requirements? {#adea-ai-native-skill-requirements} **Short answer.** It depends, but age risk is concrete when a screen rewards prior AI exposure rather than present ability or role-specific judgment. Age claims sit on a different doctrinal branch. The Supreme Court recognized disparate-impact claims under the ADEA in *Smith v. City of Jackson*, 544 U.S. 228 (2005), and *Meacham v. Knolls Atomic Power Laboratory*, 554 U.S. 84 (2008), then made clear that the employer bears the burden on the `reasonable factor other than age` defense. [^smith-v-city-of-jackson-544-u-s-228-2005][^meacham-v-knolls-atomic-power-laboratory-554-u-s] The current regulation says an employment practice harming older workers is unlawful unless justified by a "reasonable factor other than age"[^29-c-f-r-section-1625-7], and it expressly asks whether the factor was tied to a legitimate business purpose, applied fairly, and evaluated for adverse impact. [^29-c-f-r-section-1625-7] Holland & Knight's Workday alert matters because it moves the age piece out of theory. The case is not about AI fluency or prompt engineering. It is about AI-driven applicant screening. But the point is still useful: a federal court allowed a nationwide ADEA collective to proceed where older applicants alleged the AI system disproportionately harmed them. [^holland-knight-federal-court-allows-collective-a] That makes it harder to dismiss age-based statistical claims around newer AI screens as speculative. The age issue is the clearest one in the public data. Pew reported in June 2025 that 58% of adults under 30 had used ChatGPT, versus 10% of adults 65 and older. The same report said 38% of employed adults ages 18 to 29 had used ChatGPT at work, compared with 18% of workers 50 and older. [^pew-research-center-34-of-u-s-adults-have-used-c] NORC's May 2025 workplace survey is starker: 74% of workers age 60+ reported never using AI at work. [^norc-amerispeak-ai-adoption-report-may-2025] That means a `must already use AI` requirement often sorts for prior exposure before it sorts for present ability. Sex-based disparity is the next obvious pressure point. NORC's May 2025 report found daily workplace AI use at 20% for men and 10% for women, with 66% of women reporting no AI use at work versus 52% of men. [^norc-amerispeak-ai-adoption-report-may-2025] By Q3 2025, NORC said women had narrowed the gap, with equal daily workplace use but still higher non-use. [^norc-ai-adoption-report-tracking-the-rise-of-ai] The Harvard working paper is more structural: across 18 studies, women had about 22% lower odds of using generative AI tools than men, and U.S. representative samples often showed a 10 to 20 percentage point gap. [^harvard-business-school-working-paper-global-evi] So the right read is not that the gender issue is fixed or permanent. It is that the gap is moving, but still large enough to make blanket fluency gates a foreseeable Title VII issue. Perhaps a new role built specifically around model oversight or AI workflow design gets more room under the ADEA than a retrofitted generalist role. But the case law on AI-era skill gates is not there yet. [^29-c-f-r-section-1625-7][^smith-v-city-of-jackson-544-u-s-228-2005][^meacham-v-knolls-atomic-power-laboratory-554-u-s] ## Can AI-skill requirements create national-origin disparate-impact claims for employers? {#national-origin-ai-skill-requirements} **Short answer.** Unclear, because Title VII permits national-origin disparate-impact theories, but public workplace AI-use data are thinner than for age and sex. Race and national-origin exposure is harder to generalize from public surveys alone. Pew's September 2025 research found Asian adults reporting higher AI awareness and interaction than White, Hispanic, and Black adults. [^pew-research-center-americans-awareness-of-ai-an] NORC's 2025 data, though, do not tell one clean exclusion story across workplace use. Some groups lead in one measure and trail in another. [^norc-amerispeak-ai-adoption-report-may-2025-2][^norc-ai-adoption-report-tracking-the-rise-of-ai-2] Title VII plainly allows national-origin disparate-impact theories, and EEOC guidance says national origin includes `physical, cultural, or linguistic characteristics` and can reach practices that disproportionately affect people on that basis. [^eeoc-enforcement-guidance-on-national-origin-dis] But the current public U.S. evidence base is thinner here than it is on age and sex. That makes actual applicant-flow or employee-selection data more important than generalized `AI divide` rhetoric. Perhaps farther than current public discussion suggests, because EEOC guidance expressly reaches linguistic characteristics and practices with disproportionate impact on national origin. But the public U.S. survey record on workplace AI use by national origin or language is still thin enough that broad claims here would outrun the evidence. [^eeoc-enforcement-guidance-on-national-origin-dis] ## What records should employers keep for an AI-skill hiring screen? {#records-for-ai-skill-screen} **Short answer.** Usually, employers should keep applicant-flow and component-level selection data, annual adverse-impact analyses, and validity evidence for each AI-skill screen. Recordkeeping is where the doctrine becomes practical. UGESP Section 1607.15 requires users to keep, for each job, information on whether the total selection process has adverse impact, to make adverse-impact determinations at least annually, and, where the total process does show adverse impact, to keep component-level records showing which parts of the process caused it and what validity evidence exists. [^29-c-f-r-section-1607-15] The Guidelines also say agencies may draw an inference of adverse impact when the user has not maintained the required data. [^29-c-f-r-part-1607-2] EEOC recordkeeping rules separately require personnel and employment records, including application forms and other hiring records, to be preserved for one year and longer if a charge is filed. [^29-c-f-r-section-1602-14] So the legal story around an AI-skill screen is partly statistical and partly archival. Documentation becomes decisive very quickly. Where the company can identify the exact AI capability that matters, measure it through the actual work, compare lower-impact alternatives, and preserve component-level data, the dispute starts to look like validation. Where the record is just a recruiter preference, a vague posting, or a vendor screen nobody can explain, the requirement starts to look like a proxy for familiarity, age, education, and prior access. The Guidelines are explicit that annual impact review, component-level records, and validity evidence belong in the file when adverse impact appears, and that missing data can itself support an inference against the employer. [^29-c-f-r-part-1607-2][^29-c-f-r-section-1607-15][^29-c-f-r-section-1602-14] ## Did the 2025 federal pullback change AI-skill disparate-impact lawsuits? {#federal-pullback-ai-disparate-impact} **Short answer.** No, the 2025 federal pullback changed enforcement posture more than the underlying merits law. On the 2025 federal pullback, the firm commentary is also fairly consistent. Mayer Brown reads the April 23, 2025 executive order as a real change in federal enforcement priorities, but says it is still unclear how much effect that will have on private litigation because Title VII, the ADEA, and long-standing precedent remain in place. [^mayer-brown-trump-executive-order-seeks-to-elimi] That is probably the right synthesis for this topic too. The final consequence is about timing. The April 2025 executive order changed federal appetite for disparate-impact enforcement. It did not rewrite the statutes. So the center of gravity may have shifted away from agency-led pressure and toward private claims, state or local overlays, and disputes where the employer's own records become the main evidence. [^white-house-restoring-equality-of-opportunity-an][^mayer-brown-trump-executive-order-seeks-to-elimi][^holland-knight-federal-court-allows-collective-a-2] We think the better reading is mostly the forum. The statutory framework remains, but the relative weight of EEOC action, private suits, and state or local rules is still adjusting. [^white-house-restoring-equality-of-opportunity-an][^mayer-brown-trump-executive-order-seeks-to-elimi] [^about]: By Steven Obiajulu, J.D. Published by [openagreements.org](https://openagreements.org). Last reviewed 2026-04-20. License: CC BY 4.0. Steven Obiajulu, J.D. edits this topic article for Federal + multi-state coverage. It synthesizes legal sources and is not legal advice. This article is for informational purposes only and does not create an attorney-client relationship. [^42-u-s-c-section-2000e-2-k]: **42 U.S.C. section 2000e-2(k)** — "job related for the position in question and consistent with business necessity" *42 U.S.C. section 2000e-2(k).* [^griggs-v-duke-power-co-401-u-s-424-1971]: **Griggs v. Duke Power Co., 401 U.S. 424 (1971)** — "The Act proscribes not only overt discrimination but also practices that are fair in form, but discriminatory in operation. The touchstone is business necessity." *Griggs v. Duke Power Co., 401 U.S. 424 (1971).* [^albemarle-paper-co-v-moody-422-u-s-405-1975]: **Albemarle Paper Co. v. Moody, 422 U.S. 405 (1975)** — "It is true that backpay is not an automatic or mandatory remedy; like all other remedies under the Act, it is one which the courts ‘may’ invoke." *Albemarle Paper Co. v. Moody, 422 U.S. 405 (1975).* [^watson-v-fort-worth-bank-trust-487-u-s-977-1988]: **Watson v. Fort Worth Bank & Trust, 487 U.S. 977 (1988)** — "We conclude, accordingly, that subjective or discretionary employment practices may be analyzed under the disparate impact approach in appropriate cases." *Watson v. Fort Worth Bank & Trust, 487 U.S. 977 (1988).* [^connecticut-v-teal-457-u-s-440-1982]: **Connecticut v. Teal, 457 U.S. 440 (1982)** — "We hold that the ‘bottom line’ does not preclude respondent employees from establishing a prima facie case, nor does it provide petitioner employer with a defense to such a case." *Connecticut v. Teal, 457 U.S. 440 (1982).* [^eeoc-questions-and-answers-to-clarify-and-provid]: **EEOC, Questions and Answers to Clarify and Provide a Common Interpretation of...** — "all selection procedures used to make employment decisions" *EEOC, Questions and Answers to Clarify and Provide a Common Interpretation of the Uniform Guidelines on Employee Selection Procedures.* [^29-c-f-r-part-1607]: **29 C.F.R. Part 1607** — "The use of any selection procedure which has an adverse impact on the hiring, promotion, or other employment or membership opportunities of members of any race, sex, or ethnic group will be considered to be discriminatory and inconsistent with these guidelines" *29 C.F.R. Part 1607.* [^eeoc-what-is-the-eeoc-s-role-in-ai]: **EEOC, What is the EEOC's role in AI?** — "These laws apply to the use of AI and other new technologies in employment just as they apply to other employment practices." *EEOC, What is the EEOC's role in AI?.* [^eeoc-employment-discrimination-and-ai-for-worker]: **EEOC, Employment Discrimination and AI for Workers** — "Federal employment discrimination laws protect you when AI systems are used to discriminate against you on the basis of your race, color, religion, sex (including gender, sexual orientation, and pregnancy), national origin, age (40 or older), disability, or genetic information." *EEOC, Employment Discrimination and AI for Workers.* [^littler-eeoc-issues-guidance-on-use-of-artificia]: **Littler commentary** — "EEOC technical assistance documents are not voted upon or otherwise approved by the full Commission, and are not intended to create new policy, but rather to apply existing law and Commission policy to new or specific fact patterns." *Littler, EEOC Issues Guidance on Use of Artificial Intelligence Tools in Employment Selection Procedures Under Title VII.* [^mayer-brown-eeoc-issues-title-vii-guidance-on-em]: **Mayer Brown commentary** — "The EEOC’s AI Disparate Impact Guidance makes clear that the EEOC treats employer use of algorithmic decision-making tools as an employment ‘selection procedure’ under Title VII." *Mayer Brown, EEOC Issues Title VII Guidance on Employer Use of AI, Other Algorithmic Decision-Making Tools.* [^fisher-phillips-eeoc-s-latest-ai-guidance-sends]: **Fisher Phillips commentary** — "an improper application of AI could violate Title VII, the federal anti-discrimination law, when used for recruitment, hiring, retention, promotion, transfer, performance monitoring, demotion, or dismissal." *Fisher Phillips, EEOC's Latest AI Guidance Sends Warning to Employers: 5 Things You Need to Know.* [^mintz-ai-in-the-workplace-issue-spotting-for-emp]: **Mintz commentary** — "At the same time, existing federal, state, and local anti-discrimination statutes – including Title VII, the ADEA, and the ADA – apply with full force to AI-assisted decision-making" *Mintz, AI in the Workplace: Issue Spotting for Employers.* [^the-u-s-department-of-labor-s-artificial-intelli]: **The U.S. Department of Labor's Artificial Intelligence Literacy Framework** — "every worker will need baseline AI literacy skills" *The U.S. Department of Labor's Artificial Intelligence Literacy Framework.* [^eeoc-employment-tests-and-selection-procedures]: **EEOC, Employment Tests and Selection Procedures** — "Use of tests and other selection procedures can also violate the federal anti-discrimination laws if they disproportionately exclude people in a particular group by race, sex, or another covered basis, unless the employer can justify the test or procedure under the law." *EEOC, Employment Tests and Selection Procedures.* [^42-u-s-c-section-2000e-2-k-2]: **42 U.S.C. section 2000e-2(k)** — "job related for the position in question and consistent with business necessity" *42 U.S.C. section 2000e-2(k).* [^smith-v-city-of-jackson-544-u-s-228-2005]: **Smith v. City of Jackson, 544 U.S. 228 (2005)** — "the RFOA provision plays its principal role by precluding liability if the adverse impact was attributable to. a nonage factor that was ‘reasonable.’" *Smith v. City of Jackson, 544 U.S. 228 (2005).* [^meacham-v-knolls-atomic-power-laboratory-554-u-s]: **Meacham v. Knolls Atomic Power Laboratory, 554 U.S. 84 (2008)** — "The question is whether an employer facing a disparate-impact claim and planning to defend on the basis of RFOA must not only produce evidence raising the defense, but also persuade the factfinder of its merit. We hold that the employer must do both." *Meacham v. Knolls Atomic Power Laboratory, 554 U.S. 84 (2008).* [^29-c-f-r-section-1625-7]: **29 C.F.R. section 1625.7** — "reasonable factor other than age" *29 C.F.R. section 1625.7.* [^holland-knight-federal-court-allows-collective-a]: **Holland & Knight commentary** — "The court, by denying Workday's motion to dismiss, recognized Mobley's claim as plausible under the Age Discrimination in Employment Act (ADEA), based on a disparate impact theory." *Holland & Knight, Federal Court Allows Collective Action Lawsuit Over Alleged AI Hiring Bias.* [^pew-research-center-34-of-u-s-adults-have-used-c]: **Pew Research Center, 34% of U.S. adults have used ChatGPT, about double the share in 2023** — "34% of U.S. adults say they have ever used ChatGPT, according to a Pew Research Center survey." *Pew Research Center, 34% of U.S. adults have used ChatGPT, about double the share in 2023.* [^norc-amerispeak-ai-adoption-report-may-2025]: **NORC, AmeriSpeak: AI Adoption Report (May 2025)** — "Fifteen percent of employed Americans reported using AI at the workplace at least once a day, 11% use it at least once a week, and 15% use it less often. More than half of employed Americans (58%) never use AI at work." *NORC, AmeriSpeak: AI Adoption Report (May 2025).* [^norc-ai-adoption-report-tracking-the-rise-of-ai]: **NORC, AI Adoption Report: Tracking the Rise of AI in American Lives | Q3 2025** — "AI is becoming a standard tool for both Americans’ personal tasks and professional workflows." *NORC, AI Adoption Report: Tracking the Rise of AI in American Lives | Q3 2025.* [^harvard-business-school-working-paper-global-evi]: **Harvard Business School Working Paper, Global Evidence on Gender Gaps and Generative AI** — "In this paper, we show that recently identified gender gaps in generative AI use are nearly universal." *Harvard Business School Working Paper, Global Evidence on Gender Gaps and Generative AI.* [^pew-research-center-americans-awareness-of-ai-an]: **Pew Research Center, Americans' awareness of AI and views of use in daily life, control over it** — "95% of U.S. adults say they have heard at least a little about artificial intelligence." *Pew Research Center, Americans' awareness of AI and views of use in daily life, control over it.* [^norc-amerispeak-ai-adoption-report-may-2025-2]: **NORC, AmeriSpeak: AI Adoption Report (May 2025)** — "Fifteen percent of employed Americans reported using AI at the workplace at least once a day, 11% use it at least once a week, and 15% use it less often. More than half of employed Americans (58%) never use AI at work." *NORC, AmeriSpeak: AI Adoption Report (May 2025).* [^norc-ai-adoption-report-tracking-the-rise-of-ai-2]: **NORC, AI Adoption Report: Tracking the Rise of AI in American Lives | Q3 2025** — "AI is becoming a standard tool for both Americans’ personal tasks and professional workflows." *NORC, AI Adoption Report: Tracking the Rise of AI in American Lives | Q3 2025.* [^eeoc-enforcement-guidance-on-national-origin-dis]: **EEOC Enforcement Guidance on National Origin Discrimination** — "National origin discrimination means discrimination because an individual (or his or her ancestors) is from a certain place or has the physical, cultural, or linguistic characteristics of a particular national origin group." *EEOC Enforcement Guidance on National Origin Discrimination.* [^29-c-f-r-section-1607-15]: **29 C.F.R. section 1607.15** — "Users of selection procedures other than those users complying with section 15A(1) below should maintain and have available for each job information on adverse impact of the selection process for that job and, where it is determined a selection process has an adverse impact, evidence of validity as set forth below." *29 C.F.R. section 1607.15.* [^29-c-f-r-part-1607-2]: **29 C.F.R. Part 1607** — "The use of any selection procedure which has an adverse impact on the hiring, promotion, or other employment or membership opportunities of members of any race, sex, or ethnic group will be considered to be discriminatory and inconsistent with these guidelines" *29 C.F.R. Part 1607.* [^29-c-f-r-section-1602-14]: **29 C.F.R. section 1602.14** — "In the case of involuntary termination of an employee, the personnel records of the individual terminated shall be kept for a period of one year from the date of termination." *29 C.F.R. section 1602.14.* [^mayer-brown-trump-executive-order-seeks-to-elimi]: **Mayer Brown commentary** — "The Order announces the Administration’s intent to ‘seek to eliminate the use of disparate-impact liability in all contexts to the maximum degree possible to avoid violating the Constitution, Federal civil rights laws, and basic American ideals.’" *Mayer Brown, Trump Executive Order Seeks to Eliminate Disparate-Impact Liability.* [^white-house-restoring-equality-of-opportunity-an]: **White House, Restoring Equality of Opportunity and Meritocracy** — "It is the policy of the United States to eliminate the use of disparate-impact liability in all contexts to the maximum degree possible to avoid violating the Constitution, Federal civil rights laws, and basic American ideals." *White House, Restoring Equality of Opportunity and Meritocracy.* [^holland-knight-federal-court-allows-collective-a-2]: **Holland & Knight commentary** — "The court, by denying Workday's motion to dismiss, recognized Mobley's claim as plausible under the Age Discrimination in Employment Act (ADEA), based on a disparate impact theory." *Holland & Knight, Federal Court Allows Collective Action Lawsuit Over Alleged AI Hiring Bias.*