Trump’s Executive Order: Fundamentally Misunderstanding the Law
On April 23, 2025, President Trump signed Executive Order 14281, Restoring Equality of Opportunity and Meritocracy.[1] Its stated goal is “to eliminate the use of disparate-impact liability in all contexts to the maximum degree possible.”[2]
The order directed enforcement agencies such as DOJ, EEOC, FTC, and the Consumer Financial Protection Bureau (CFPB) to “deprioritize” enforcement of disparate impact laws, pushing them not just to drop pending cases but also to ask courts to lift consent decrees and injunctions won based on the theory.[3] It also revoked presidential approval for disparate impact regulations under Title VI—the main authority to prevent discriminatory uses of federal funds[4]—and directed agencies to formally rescind them.[5] On December 10, 2025, without first publishing a proposed rule or seeking public comment, DOJ issued an immediately effective rule eliminating Title VI’s disparate impact provisions, which had governed recipients of federal funding for over half a century.[6] The CFPB has also issued a proposed rule to eliminate disparate impact from regulations implementing the Equal Credit Opportunity Act.[7]
Finally, Trump’s April 2025 order directed the Attorney General to determine “whether any Federal authorities preempt” state-level disparate impact liability and whether such state laws “have constitutional infirmities that warrant Federal action.”[8] In December 2025, he directed the Attorney General to create an “AI Litigation Task Force” to pursue lawsuits challenging state-level regulation of AI, and ordered other agencies to take steps to shore up the tenuous case for preemption.[9] These actions make clear that the administration does not merely intend to shirk its duty to enforce federal anti-discrimination law. It intends to interfere with state laws, as well.[10]
Trump’s attack on disparate impact rests on at least three grievous legal errors.
- First, the order wrongly asserts that under disparate impact liability, “a near insurmountable presumption of unlawful discrimination exists where there are any differences in outcomes in certain circumstances among different races, sexes, or similar groups, . . . even if everyone has an equal opportunity to succeed.” The first half of this sentence is a misstatement of the law. The second half misunderstands equal opportunity.
Disparate impact does not turn on “any differences.” In fact, under Supreme Court precedent and statutory law, plaintiffs must show that the challenged practice causes a “significant” or “substantial” statistical disparity to avoid having their case immediately dismissed.[11] That is just to make out a prima facie case, meaning showing that the claim appears to have merit on its face. At the second step of the analysis, companies acting in good faith typically do not have a problem establishing that the challenged practice is consistent with business necessity. Then the burden shifts back to the plaintiffs to identify a less discriminatory alternative. Proving the existence of a viable alternative can be complicated and expensive.
If plaintiffs manage to prove all of this, they have proved that the challenged practice unnecessarily harmed them based on race. They have proved that they didn’t have “an equal opportunity to succeed.” Put differently, they have proved discrimination.[12]
- Second, the order wrongly asserts that disparate impact requires defendants to “engage in racial balancing.” Changing a practice that systematically harms people of a certain race—when the practice doesn’t actually serve a company’s legitimate interests, or when those interests can be advanced by a less discriminatory practice—is not racial balancing. It is removing an indefensible source of bias. The result is a fairer process for everyone. It is true that the analysis involves doing some math, but as noted, disparate impact doctrine does not demand parity. It allows considerable variation in outcomes before applying any scrutiny at all. In addition, Title VII itself explicitly bans quotas.[13]
It bears noting that outside of zero-sum contexts like hiring, decisionmakers are not making choices between two candidates. They may be assessing the value of a family’s home or predicting a patient’s cancer risk. Disparate impact law remains relevant to ensure they do not use tools or processes that skew results based on race or other irrelevant traits, and the concept of quotas has no plausible applicability.
- Third, the order wrongly asserts that disparate impact is unconstitutional. Its claim that the doctrine “runs contrary to equal protection under the law” has its roots in a concurrence by Justice Antonin Scalia in Ricci v. DeStefano,[14] but ignores subsequent Supreme Court decisions.
In Ricci, a group of New Haven, Connecticut firefighters sued the city for intentional race discrimination under Title VII. The city had discarded the results of a promotion exam after seeing that almost all those who scored high enough were White and none of them were Black. The city explained that it scrapped the results because it feared being sued under Title VII for disparate impact. But the Court found that New Haven did not have a foundation to conclude the test was discriminatory. Specifically, city officials did not thoroughly evaluate whether the test was job-related and consistent with business necessity, or whether there existed any less discriminatory alternative that served its needs. The city therefore lacked “a strong basis in evidence” to believe it could be held liable under disparate impact for certifying the test results. Under those circumstances, the Court found that its decision not to use the results amounted to disparate treatment of the firefighters who had passed the test (i.e., declining to promote them because of their race).[15]
Scalia went further, suggesting that disparate impact liability itself might be unlawful under the Constitution. He contended that by “requiring employers to evaluate the racial outcomes of their policies, and to make decisions based on (because of) those racial outcomes,” disparate impact forces employers to engage in discriminatory “racial decisionmaking.”[16]
On this reasoning, any attempt to prevent racially disparate consequences—no matter how traceable to historical discrimination, how grounded in arbitrary considerations, how predictably unjust, or how easy to avoid while still serving an employer’s legitimate interests—would itself be racial discrimination.
This is not the law.[17] In a pair of post-Ricci cases about University of Texas admissions, not a single Justice questioned the state’s “Top Ten Percent” plan—under which the university automatically admitted the top 10% of each Texas high school—even though its purpose was to increase diversity.[18] As Professor Reva Siegel has observed, although the plan was race-conscious in its purpose of creating equal opportunity for students of color, it was constitutional because it was race-neutral in form and did not classify students based on race. The same is true of disparate impact. The doctrine is race-conscious in that it aims to avert unjustified adverse impacts based on race. But it does not require a decisionmaker to use race as a selection criterion, and it does not advantage or disadvantage any group based on race.[19] In this sense, it is wholly distinct from affirmative action.
Moreover, in 2015 the Supreme Court upheld the existence of disparate impact under the Fair Housing Act in Texas Department of Housing & Community Affairs v. Inclusive Communities Project, Inc.[20] Constitutional arguments abounded in the case, with several amicus briefs arguing that disparate impact law is unconstitutional. The Court was not persuaded.[21] Writing for the majority, Justice Anthony Kennedy explained that “disparate-impact liability has always been properly limited in key respects that avoid the serious constitutional questions that might arise,” such as by holding that statistical disparity alone is insufficient to establish liability.[22]
Doubtless, President Trump’s order sets back the cause of non-discrimination. It does not change the law, however. Disparate impact remains prohibited under federal statutes—enforceable by state attorneys general and private parties—covering employment, housing, lending, and other spheres of life. Several state laws likewise impose disparate impact liability,[23] and some states have specifically targeted AI systems that have discriminatory effects.[24] The Trump administration’s arguments that these statutes may be preempted by or violate federal law are weak and unlikely to prevail.[25]
[1] President Donald J. Trump, Executive Order 14281, Restoring Equality of Opportunity and Meritocracy, 90 Fed. Reg. 17537 (Apr. 23, 2025), https://www.federalregister.gov/documents/2025/04/28/2025-07378/restoring-equality-of-opportunity-and-meritocracy.
[2] Id.
[3] Agencies had already removed key guidance documents from their websites in the early days of the Trump Administration. See, e.g., EEOC, Select Issues: Assessing Adverse Impact in Software, Algorithms, and Artificial Intelligence Used in Employment Selection Procedures Under Title VII of the Civil Rights Act of 1964 (May 18, 2023), available at https://data.aclum.org/storage/2025/01/EOCC_www_eeoc_gov_laws_guidance_select-issues-assessing-adverse-impact-software-algorithms-and-artificial.pdf; U.S. Department of Housing and Urban Development, Guidance on Application of the Fair Housing Act to the Screening of Applicants for Rental Housing (April 29, 2024), available at https://www.fairhousingnc.org/wp-content/uploads/2024/08/FHEO_Guidance_on_Screening_of_Applicants_for_Rental_Housing.pdf.
[4] See Alexander v. Sandoval, 532 U.S. 275 (2001) (holding that Title VI’s statutory prohibition on discrimination, Section 601, prohibits only intentional discrimination, and that there is no private right of action to enforce disparate impact regulations promulgated under Section 602, meaning only the federal government can enforce them).
[5] Executive Order 14281.
[6] Final Rule, Rescinding Portions of Department of Justice Title VI Regulations, supra note 7.
[7] 90 Fed. Reg. 50901 (Nov. 13, 2025), https://www.federalregister.gov/documents/2025/11/13/2025-19864/equal-credit-opportunity-act-regulation-b.
[8] Id.
[9] Executive Order 14365, supra note 6.
[10] Trump issued another order about AI that also warrants comment. Executive Order 14319, Preventing Woke AI in the Federal Government, announced that the federal government—the world’s largest buyer—would only purchase generative AI systems developed in accordance with “ideological neutrality.” 90 Fed. Reg. 35389 (July 23, 2025), https://www.federalregister.gov/documents/2025/07/28/2025-14217/preventing-woke-ai-in-the-federal-government. By way of definition, the order specifies that large language models must not “encode” diversity, equity, and inclusion in their outputs. Id. Technologists have rightly pointed out that this mandate “positions one ideological perspective as the default standard for neutrality,” and “efforts to align models” with it “risk introducing new distortions.” Amy Winecoff & Chinmay Deshpande, Center for Democracy & Technology, Anti-Woke AI Is a Technical Mirage (Aug. 8, 2025), https://cdt.org/insights/anti-woke-ai-is-a-technical-mirage. Indeed, by preventing developers from addressing the known biases discussed in this paper, Trump’s “woke AI” order may actually require discriminatory design as a condition of AI vendors obtaining federal contracts.
[11] Albemarle, 422 U.S. at 405 (a plaintiff must show that employment tests “select applicants for hire or promotion in a racial pattern significantly different from that of the pool of applicants”); 29 C.F.R. § 1607.16(Q) (defining “adverse impact” as a “substantially different rate of selection in hiring, promotion or other employment decision which works to the disadvantage of members of a race, sex, or ethnic group”).
[12] See EEO Leaders’ Statement on Disparate Impact, President Trump’s Executive Order on Disparate Impact Analysis Is Legally Incorrect and Will Undermine Meritocracy and Equal Employment Opportunity (May 2025), https://bit.ly/3F7A6bh (“[T]he entire concept of disparate impact is that unjustified and significant differences in
outcome resulting from a ‘neutral’ policy means that people of different races or sexes are not being given an equal opportunity to succeed.”).
[13] 42 U.S.C. § 2000e-2(j).
[14] 557 U.S. 557 (2009).
[15] Id.
[16] Id. at 594 (Scalia, J., concurring).
[17] Zachary Best & Stephen Hayes, Executive Order on Disparate Impact: An Explainer, 3 (May 9, 2025) (“No court has ever held that disparate impact runs afoul of the Constitution.”), https://www.relmanlaw.com/media/cases/1965_Executive%20Order%20on%20Disparate%20Impact%20Explainer.pdf.
[18] Fisher v. Univ. of Texas at Austin (Fisher I), 570 U.S. 297 (2013); Fisher v. Univ. of Texas at Austin (Fisher II), 579 U.S. 365 (2016). See also id. at 532 (Thomas, J., dissenting) (describing the state law establishing the Top Ten Percent plan as “facially race-neutral law” that “served to equalize competition between students who live in relatively affluent areas with superior schools and students in poorer areas” and “tended to benefit African-American and Hispanic students, who are often trapped in inferior public schools”).
[19] Reva Siegel, Race-Conscious but Race-Neutral: The Constitutionality of Disparate Impact in the Roberts Court, 66 Ala. L. Rev. 653, 672-78 (2013). Justice Kennedy, the author of both Fisher opinions, had previously explained that policymakers could pursue race-conscious goals through race-neutral means. See Parents Involved in Cmty. Schs. v. Seattle Sch. Dist. No. 1, 551 U.S. 701, 789 (2006) (Kennedy, J., concurring in part and concurring in the judgment) (“School boards may pursue the goal of bringing together students of diverse backgrounds and races through other means, including strategic site selection of new schools; drawing attendance zones with general recognition of the demographics of neighborhoods; allocating resources for special programs; recruiting students and faculty in a targeted fashion; and tracking enrollments, performance, and other statistics by race. These mechanisms are race conscious but do not lead to different treatment based on a classification that tells each student he or she is to be defined by race, so it is unlikely any of them would demand strict scrutiny to be found permissible.”). Indeed, Justice Scalia himself had acknowledged as much 20 years before Ricci. See City of Richmond v. J.A. Croson, 488 U.S. 469, 526 (1989) (Scalia, J., concurring in the judgment) (“A State can, of course, act to undo the effects of past discrimination in many permissible ways that do not involve classification by race.”) (internal quotation marks omitted).
[20] Tex. Dep’t of Hous. & Cmty. Affairs v. Inclusive Cmtys. Project, Inc., 576 U.S. 519 (2015).
[21] See Samuel R. Bagenstos, Disparate Impact and the Role of Classification and Motivation in Equal Protection Law after Inclusive Communities, 101 Cornell L. Rev. 1115, 1127-28 (2016) (“Because the Fair Housing Act does not expressly provide for disparate-impact liability, if a majority of the Court had serious constitutional concerns about disparate impact claims per se, the Court would likely have avoided the constitutional problem by reading the statute not to provide for such claims. By holding that the Fair Housing Act does provide for disparate-impact liability, the Court must therefore have rejected the argument that disparate impact law is unconstitutional.”).
[22] Inclusive Cmtys., 576 U.S. at 536-37.
[23] See, e.g., California Fair Employment and Housing Act, Cal. Gov’t Code § 12955.8(b); Colorado Anti-Discrimination Act, C.R.S. §§ 24-34-402, 24-34-502; Illinois Human Rights Act, 775 I.L.C.S. 5/2-102; Mass. Gen. Laws ch. 151B § 4; N.J.S.A. §§ 13:13-2.5, 13:13-3.4(f)(2), 13:13-4.11; Washington Law Against Discrimination, R.C.W. § 49.60.
[24] See, e.g., Colorado SB 24-205, Consumer Protections for Artificial Intelligence, codified at C.R.S. § 6-1-1701 et seq. (2024), https://leg.colorado.gov/bills/sb24-205; Illinois H.B. 3773, amending the Illinois Human Rights Act (2024), https://legiscan.com/IL/bill/HB3773/2023; New Jersey Attorney General, Division on Civil Rights, Guidance on Algorithmic Discrimination and the New Jersey Law Against Discrimination (2025), https://www.nj.gov/oag/newsreleases25/2025-0108_DCR-Guidance-on-Algorithmic-Discrimination.pdf.
[25] See Charlie Bullock, supra note 8 (stating that the case for preemption under the Commerce Clause is “legally dubious and unlikely to succeed in court); Gibson Dunn, President Trump’s Latest Executive Order on AI Seeks to Preempt State Laws (Dec. 15, 2025), https://www.gibsondunn.com/president-trump-latest-executive-order-on-ai-seeks-to-preempt-state-laws (explaining why DOJ’s preemption arguments “are unlikely to be successful” and why the contemplated FCC and FTC actions would not be a basis for preemption).