TL;DR: On August 7, 2026, the FTC voted 2-0 to stop pursuing disparate-impact discrimination claims under Section 5 and ECOA, following an EEOC guidance withdrawal, a new EEOC enforcement plan, and a DOJ legal opinion that all point the same direction. None of that touches Title VII's disparate-impact doctrine itself, NYC Local Law 144, Illinois AI employment law, or California's ADMT rules, and it does not stop private lawsuits. If you use AI in hiring, your bias-audit obligations have not gone away, only the federal agency that used to chase them for you has.
What the FTC actually did on August 7
The Federal Trade Commission voted 2-0 to approve a formal Disparate-Impact Policy Statement. The document says the Commission will no longer bring claims under Section 5 of the FTC Act, or the Equal Credit Opportunity Act, based on a disparate-impact theory, meaning a neutral practice that produces a statistically unequal outcome across demographic groups even without proof that anyone meant to discriminate.
FTC Chairman Andrew Ferguson did not hedge. "Disparate-impact claims are nearly impossible to square with our colorblind Constitution," he said. "They impose liability for discrimination without any evidence that anyone intended to discriminate." He went further: "The commission never had authority to impose disparate-impact liability. Today, we announce that the commission will never do so again."
The Commission's own account announced the statement the same day.
https://x.com/FTC/status/2085716144724300154
That is a bright line, not a nuance. The Commission's stated position now is that a company cannot be found liable under Section 5 or ECOA purely because an algorithm, a lending model, or a hiring tool produced skewed outcomes. Intent, or something close to it, is now the Commission's threshold.
The policy statement was not purely theoretical. The FTC used the same announcement to revisit compliance obligations tied to earlier settlements with three auto dealer groups, Napleton Inc., Passport Auto Group, and Coulter Motor Co. LLC, which had been bound by statistical disparate-impact monitoring requirements from prior consent orders. None of those cases involved AI. They involved financing and markup practices at car dealerships. But they are the clearest evidence yet that this is not a symbolic statement sitting on a shelf, it is already changing what the FTC requires of companies it has already settled with.
One distinction the Commission was careful to preserve: disparate-treatment claims, meaning intentional discrimination, are still fully in play under ECOA and other fair-lending statutes. Dealerships, lenders, and by extension AI vendors, still have to avoid deliberately steering outcomes by protected class. What is gone is liability based purely on statistical outcome gaps.
This was not a one-agency decision
If the FTC's move feels sudden, it is worth seeing it as the fourth step in a sequence that has been running since early 2025, not a standalone event.
January 2025, the EEOC removed its May 2023 Technical Assistance document on AI and Title VII from its website. That guidance had told employers to test AI hiring tools for adverse impact against protected groups. Its removal did not repeal the underlying law, but it took away the agency's own how-to guide.
April 23, 2025, the White House issued an executive order titled "Restoring Equality of Opportunity and Meritocracy," directing federal agencies, by name including the FTC, to evaluate and unwind enforcement actions built on disparate-impact theory wherever legally possible.
June 4, 2026, EEOC Chair Andrea Lucas signed a new National Enforcement Plan for fiscal years 2025 through 2029, replacing the prior administration's Strategic Enforcement Plan and explicitly deprioritizing disparate-impact theories in EEOC investigations and litigation.
June 9, 2026, the Department of Justice's Office of Legal Counsel issued a memorandum arguing that disparate-impact liability, as applied under federal civil rights statutes, conflicts with equal protection principles and should be read narrowly going forward.
August 7, 2026, the FTC's policy statement completed the set. Three agencies that all touch AI hiring and lending tools in some way, the FTC through Section 5 and ECOA, the EEOC through Title VII, and DOJ through its legal interpretation authority, are now aligned against pursuing disparate-impact theories at the federal level.
That is a real, coordinated shift in federal enforcement posture. It is not, on its own, a change in what the underlying laws say.
What did not change: the statute is still the statute
Here is the distinction that gets lost in headlines about the FTC "ending AI bias enforcement." Federal agencies choosing not to bring a certain type of case is an enforcement priority decision. It is not the same as a court striking down disparate-impact doctrine, and it is not the same as Congress repealing Title VII.
Title VII's disparate-impact doctrine traces back to Griggs v. Duke Power Co., a 1971 Supreme Court decision holding that a facially neutral employment practice can violate the Civil Rights Act if it disproportionately screens out a protected group and the employer cannot show the practice is job-related and consistent with business necessity. That precedent has not been overturned. The EEOC choosing not to bring cases under it does not remove a private plaintiff's right to sue under it directly.
This matters concretely because private hiring-discrimination litigation involving AI tools is already active. The ongoing Workday lawsuit, in which rejected job applicants allege the company's AI-powered screening tools disproportionately filtered out candidates over 40, Black candidates, and candidates with disabilities, is proceeding as a class action independent of what the EEOC or FTC choose to prioritize. Agencies stepping back from a legal theory does not close the courthouse door on the people that theory was designed to protect. It just means they have to walk through it themselves, or with a plaintiff's firm, instead of having the EEOC or FTC walk in front of them.
For a small team running any part of hiring through an algorithmic tool, that is the risk that survives every one of the four federal actions above: an applicant or a plaintiff's attorney with statistics showing your tool disproportionately screens out a protected group, and a Title VII claim that does not require the EEOC's cooperation to file.
What did not change: state and city law
The FTC's authority comes from federal statute. It has no power to preempt independent state or local laws that impose their own bias-testing requirements, and the August 7 statement does not attempt to.
NYC Local Law 144 still requires any employer using an automated employment decision tool on a candidate residing in New York City to commission an independent annual bias audit and publish the results before using the tool. Enforcement has gotten stricter through 2026, not looser, and it has nothing to do with FTC Section 5 authority.
Illinois still requires written notice and consent before AI-based video interview analysis, and separately requires disclosure when AI is used in employment decisions under its state employment law. Neither provision depends on federal disparate-impact theory.
California's Civil Rights Council regulations on automated decision-making technology impose their own documentation, notice, and adverse-impact assessment obligations on employers using ADMTs, built entirely on state FEHA authority, not federal law the FTC or EEOC administer.
Colorado is the one place this gets genuinely confusing, and it is worth being precise about why. Colorado's original AI Act, SB 24-205, did include an algorithmic-discrimination duty of care that resembled disparate-impact obligations. But that law's enforcement was suspended by a federal court in April 2026 after xAI sued and the Department of Justice intervened, not because of the FTC's disparate-impact policy statement. The legislature then replaced it with SB 26-189, which drops the impact-assessment framework entirely in favor of a narrower notice-and-disclosure model, effective January 1, 2027, and Colorado's attorney general has said the state will not enforce it until rulemaking finishes. If you operate in Colorado, your disparate-impact exposure there was already fading for reasons that predate and have nothing to do with the FTC. Do not read that as evidence the FTC's move caused it, and do not assume the same court fight applies to New York, Illinois, or California, which have not been challenged on these grounds.
The pattern across all four states: none of them borrow FTC Section 5 or ECOA authority to operate. They stand on independent state or local legal foundations that the FTC's policy statement cannot touch, no matter how the federal posture shifts.
The checklist for this week
1. Do not update your AI hiring compliance program based on the FTC statement alone. If your bias-audit process was built around NYC Local Law 144, Illinois notice requirements, or California ADMT rules, none of those obligations changed on August 7. Treat the FTC statement as a federal enforcement-posture update, not a green light.
2. Confirm which jurisdictions your hiring actually touches. If you have even one remote candidate or employee based in New York City, Local Law 144 can apply to your hiring tool regardless of where your company is headquartered. Map every state and city where candidates reside, not just where your office sits.
3. Keep your vendor bias-audit documentation current regardless of federal posture. Ask your AI hiring vendor for their most recent independent bias audit results and the demographic groups tested. Private plaintiffs' attorneys, not just regulators, can request this in discovery.
4. Watch the Workday case and similar private litigation, not just agency headlines. Federal agency posture and courtroom outcomes move independently now. A case like Workday's can set practical precedent for AI hiring liability even while the EEOC and FTC step back from bringing similar claims themselves.
5. Do not treat "the FTC will never do so again" as legal advice for your specific tool. That quote describes FTC enforcement policy under the current Commission. Policy statements are not binding law, can be revised by a future Commission, and do not bind state attorneys general, city agencies, or private plaintiffs at all.
6. Document today's review with a date. If you check your bias-audit obligations against this list this week, write down what you found and when. A dated record of a compliance review is worth more after the fact than a memory of having "looked into it at some point."
What to watch next
The federal disparate-impact retreat is likely not finished. Expect the Consumer Financial Protection Bureau and other financial regulators to face pressure to adopt similar language for AI-driven lending decisions, following the same executive order that shaped the FTC's move. Watch for whether any state attorney general responds to the FTC statement by explicitly reaffirming their own disparate-impact authority, the way several already have on other federal preemption arguments this year.
The more important thing to watch is not federal at all. It is whether courts hearing the Workday case, or the next AI hiring lawsuit like it, treat Title VII's disparate-impact doctrine as intact despite four federal agencies stepping back from enforcing it themselves. That answer will tell small teams more about real-world AI hiring liability than any policy statement from Washington.
Related Reading
- EEOC AI Hiring Guidance 2026: 8-Step Employer Compliance Checklist
- NYC Local Law 144 AI Bias Audit: Employer Guide 2026
- Illinois AI Employment Disclosure Law 2026
- California ADMT Employer Compliance Templates 2026
- Workday AI Lawsuit: HR Screening Compliance Checklist
- Colorado SB 26-189 Signed: What Changed and What Employers Must Do
- AI Hiring Compliance for Small Teams: The Complete Guide
