AI Regulation Tracker / For HR and People Teams
EEOC AI Hiring Rules: New Enforcement Plan and DOJ Opinion
Two federal moves in June 2026 changed how Washington treats biased hiring algorithms, but state bias-audit laws still bind employers who use them.
In June 2026 the EEOC approved a National Enforcement Plan for FY2025 to FY2029 that deprioritizes disparate-impact enforcement, the main federal theory for challenging biased AI hiring tools, and a DOJ Office of Legal Counsel opinion called the EEOC's hiring recommendations unconstitutional. Neither action repealed Title VII. Disparate-impact liability remains law, private plaintiffs can still sue, and state bias-audit rules such as NYC Local Law 144 still bind employers.
For three years, HR teams treated one federal idea as the center of gravity for AI hiring risk: disparate impact. It is the theory that says a neutral-looking screen can still break the law if it filters out protected groups at higher rates and the employer cannot show the screen is job-related. In June 2026, the federal government stepped back from that theory twice within a week. If you run resume screeners, video-interview scorers, or any automated selection tool, here is what actually changed, what did not, and what you should do on Monday.
What the EEOC actually did
On June 4, 2026, the U.S. Equal Employment Opportunity Commission approved a new National Enforcement Plan covering fiscal years 2025 through 2029. Issued under Chair Andrea R. Lucas, it rescinds and replaces the prior Strategic Enforcement Plan and took effect immediately, according to the law firm Ogletree Deakins.
The headline for HR is this line of policy: the EEOC says it will prioritize intentional discrimination, which it calls disparate treatment. According to Ogletree Deakins' summary of the plan, the agency commits to eliminate its use of disparate-impact theories to the maximum degree possible and to not commence, develop, or continue litigation that advances disparate-impact claims. The plan ties that shift to Executive Order 14281, "Restoring Equality of Opportunity and Meritocracy," which President Trump signed on April 23, 2025, directing agencies to deprioritize disparate-impact enforcement.
Why does that matter for algorithms specifically? Because disparate impact has been the federal government's primary tool for challenging biased selection systems. A vendor's resume model is not going to hand you a memo that says "screen out older applicants." The harm shows up in the pass rates. Disparate-impact analysis is how a regulator gets from "the tool looks neutral" to "the tool is illegal." Pull that theory out of the EEOC's toolkit and you have pulled out the main federal lever on AI hiring bias.
What the DOJ opinion did, and what it did not do
Days later, the picture got more attention. The Department of Justice's Office of Legal Counsel concluded that the EEOC's disparate-impact guidance under Title VII was unconstitutional, reasoning that the guidance pressured employers into race-based decisions to avoid liability, as Forbes reported on June 17, 2026. It is unusual for the DOJ to weigh in on EEOC guidance this way, which is part of why the opinion drew headlines.
Read past the headlines, though, and the legal experts quoted are blunt about the limits. "Honestly, this opinion means very little on its own," said David Glasgow, executive director of NYU's Meltzer Center for Diversity, Inclusion, and Belonging. He noted that the administration had already issued an executive order telling agencies to deprioritize disparate-impact enforcement, and that the opinion "has no binding legal effect." Workers, he added, can still file disparate-impact claims on their own.
Leslie Marant, an employment attorney and founder of The ESP Effect, made the same point from the plaintiff's side. "The law hasn't changed. Workers can still bring disparate impact claims under Title VII. Courts can still hear those claims. Congress hasn't repealed either Title VII or the Civil Rights Act of 1991." What changed, she said, is the signal: the EEOC and DOJ "appear increasingly hostile to disparate impact enforcement and will likely become less willing to investigate, pursue, or support these claims."
So be precise with your executive team. Nobody repealed the AI hiring rules. The statute is intact. What shifted is the federal posture: the agency most likely to knock on your door about a biased algorithm has told the world it would rather not.
The trap: reading a federal retreat as a free pass
Here is the mistake that will cost employers money. A retreat by two federal actors is not the same as legal safety, for three reasons.
First, private plaintiffs. Marant's own list of what disparate impact has historically been used to challenge is telling: "tests, screening tools, promotion practices, background checks, hiring criteria, and now algorithms that produce discriminatory results even when the employer insists the policy is neutral." Plaintiffs' lawyers do not need the EEOC's blessing to file. If anything, a well-documented adverse-impact pattern in your hiring tool is exactly the kind of case a private firm will take.
Second, the burden of proof is moving, not disappearing. Both attorneys quoted told workers to keep going and to document harder. That means the cases that do get filed will be built with more specificity and better evidence. An employer that assumed federal indifference and stopped testing its tools is the easiest target in that world.
Third, and most important for anyone using AI hiring tools: the states never left.
What still binds you: the state layer
Federal enforcement is one floor. It is not the whole building. Several states and cities regulate automated hiring directly, and none of them were touched by the EEOC plan or the DOJ opinion.
- New York City's Local Law 144 requires employers using automated employment decision tools to complete an independent bias audit, publish the results, and notify candidates before use. That obligation lives in city law, enforced by the city, not by the EEOC. See our breakdown of the [NYC LL144 enforcement and audit risk](/ai-regulation-news/nyc-ll144-enforcement-audit-employer-risk).
- Illinois amended its Human Rights Act to address AI in employment, adding notice duties and making it a civil-rights violation to use AI in a way that produces discriminatory effects or to use protected-class proxies like zip code. Details in our [Illinois HB 3773 notice guide](/ai-regulation-news/illinois-hb3773-ai-employment-notice).
- California adopted employment regulations covering automated decision systems under its Fair Employment and Housing Act, reaching tools used to screen or rank applicants.
State bias-audit and notice regimes generally rest on the same adverse-impact logic the EEOC is stepping away from. So the federal retreat does not shrink your exposure under them. It concentrates it there.
How to use this: what changed, what still applies, what to do
Run your talent stack through this in one sitting.
What changed:
- The EEOC has formally deprioritized disparate-impact enforcement and will not pursue disparate-impact litigation.
- A DOJ OLC opinion concluded that the EEOC's disparate-impact guidance under Title VII is unconstitutional, though the opinion is advisory only.
What still applies:
- Title VII, including disparate-impact liability, remains law. Private plaintiffs can still sue.
- NYC Local Law 144, the Illinois Human Rights Act AI provisions, California's automated-decision rules, and similar state measures still bind you.
- Your bias-audit, candidate-notice, and recordkeeping duties under those laws did not change.
What to do this quarter: 1. Inventory every automated hiring tool in use, including vendor resume screeners, video-interview scorers, chatbots, and internal scoring models. You cannot defend what you have not listed. 2. Map each tool to jurisdiction. Where do your candidates and roles sit? Pull the specific state and city rules that reach each location. 3. Confirm a current bias audit exists for any tool that needs one, and that the summary is published where the law requires. 4. Confirm candidate notice is actually firing in your applicant flow, not just written into a policy PDF. 5. Keep testing pass rates by protected group anyway. Adverse-impact data is your evidence in a private suit and your early warning before one. 6. Get your vendor's audit methodology and indemnity terms in writing. If the model produces the disparate outcome, you are the employer on the charge.
Do not use this moment to quietly retire the guardrails. Use it to make sure they are real.
The literacy gap underneath all of this
The deeper lesson for managers is that "the government relaxed the rules" is almost never the whole story, and acting on the headline version is how teams walk into liability. Someone on your side has to be able to read a regulatory shift, separate the binding part from the political signal, and translate it into an operational checklist. That is a management skill now, not just a legal one.
If your team is going to keep buying and running AI tools that make decisions about people, the people approving those tools need enough fluency to ask the right questions. That is the entire premise of learning to [use AI as a manager](/ai-for-managers): not to become a lawyer, but to stop mistaking a press release for a permission slip. Not sure where your team's gaps are? Our [two-minute quiz](/quiz) points you to the right starting point.
Frequently Asked Questions
Did the EEOC or DOJ repeal the rules against biased AI hiring tools?
No. Title VII and its disparate-impact provision are still law, and Congress has not touched them. The EEOC changed its enforcement priorities, and the DOJ issued a non-binding legal opinion. Neither one repealed anything or stopped private lawsuits.
If the EEOC will not pursue these cases, who will?
Private plaintiffs and their attorneys, plus state and city regulators. The employment lawyers quoted in the coverage explicitly told workers to keep filing and to document their evidence more carefully. Expect fewer agency investigations and, potentially, more private suits.
We only hire in states with no AI hiring law. Are we in the clear?
You are exposed to less state risk, but not zero federal risk, because private Title VII disparate-impact claims still exist nationwide. If you hire anyone in New York City, Illinois, California, or other regulated jurisdictions, those specific bias-audit and notice duties still apply to those candidates.
Is the DOJ opinion binding on courts?
No. An Office of Legal Counsel opinion is advisory. One of the legal experts in the coverage called it "an opinion from the DOJ" with "no binding legal effect." Courts decide Title VII cases under the statute and existing precedent, not under an agency opinion.
Should we cancel our scheduled bias audit to save money?
No. Bias audits are required by state and city laws that the federal changes did not affect, and the audit data is also your best evidence if a private plaintiff sues. Canceling it removes a legal requirement in some places and removes your defense everywhere.
What is the single most useful thing to do right now?
Build the inventory of every automated hiring tool you use and match each one to the laws of the places you hire. Most compliance failures start with an employer that could not say, precisely, which tools it was running and where.
Sponsored Training
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.