One AI hiring tool has rejected 1.1 billion job applications. That number came from Workday, the company that built it. It surfaced in a federal discrimination lawsuit (Mobley v. Workday, N.D. Cal.; National Law Review, March 31, 2026).

Somewhere in that pile are dyslexic candidates. A timed test or a resume scanner may have screened them out. It never saw their real skills.

The federal government used to explain when that crosses into illegal discrimination. That explanation is gone now.

The EEOC deleted its AI hiring guidance in January 2025

In January 2025, the EEOC removed its main AI hiring guidance. The page came down from eeoc.gov (Bloomberg Law, February 20, 2025).

The National Law Review checked again on March 21, 2026. The core technical guidance document still returned a 404 error. So did the companion "Tips for Workers" page (National Law Review, March 31, 2026).

The removal wasn't announced as a standalone AI decision. It was part of a wider rollback of Biden-era guidance, following Executive Order 14179 (KMK Law, February 2025).

If you search for federal AI hiring guidance today, the page is gone. Your protection under the law is not.

The law underneath it never moved

Deleting a guidance document doesn't repeal a law. Guidance just explains how existing rules apply. It doesn't create new ones.

Title I of the ADA still bans this kind of screening. It's illegal to use hiring technology that screens out a qualified disabled applicant (42 U.S.C. § 12112).

The Department of Justice publishes similar guidance of its own, for state and local government employers. That page was still live and unchanged when we checked it in July 2026 ("Algorithms, Artificial Intelligence, and Disability Discrimination in Hiring," ada.gov, May 12, 2022).

The EEOC's own Strategic Enforcement Plan still lists AI-related hiring discrimination as a priority. It runs from 2024 through 2028. Only a vote of the commissioners can change it (EEOC Strategic Enforcement Plan; Cooley LLP, February 2025).

The rulebook explaining your rights got harder to find. Your rights didn't change with it.

How AI hiring tools actually screen out dyslexia

DOJ's guidance gives real examples of this. A facial and voice analysis tool can flag speech patterns linked to a disability. That can happen even when the applicant can do the job (ada.gov, May 2022).

For a dyslexic applicant, the risk looks different. A resume scanner might penalize unusual spelling. A timed personality test might measure reading speed, not judgement.

A chatbot interview might offer no option to type instead of speak. None of these tools were built to test for dyslexia. That's exactly the problem the ADA targets.

A test measures an unrelated skill. It screens out a qualified candidate as a side effect. That's discrimination, guidance page or not.

The ADA accommodations guide covers reasonable accommodations during hiring, not only after you've got the job.

If a hiring test feels impossible to finish in time, name that before you submit it. Don't wait until you're rejected.

What to ask for before an AI tool rejects you

The ADA's accommodation duty starts when you apply. It doesn't wait until you're hired. That includes AI-driven tests and interviews (ada.gov, May 2022).

Here's the practical version. Tell the employer, in writing, that you'd like to know what tool or software is being used. Ask what help is available before you take the test. Reasonable requests include:

  • Extra time to complete a timed test or written exercise
  • A text-based interview option instead of video or voice analysis
  • A human reviewer for any automated score before a rejection is final
  • Advance notice of exactly what the test is measuring

Deciding how much to disclose at this point is its own choice. The disclosure decision guide covers timing for job applications. That's separate from disclosure after you're hired.

If you don't know what to ask for, the Job Accommodation Network keeps a free list of accommodation ideas by task, including hiring tests (askjan.org). It's a good starting point when you're not sure what to name.

Asking for an accessible hiring test is a legal right, not a favor. Put the request in writing, so there's a record if the employer refuses.

Four states didn't wait for federal guidance

While the federal page vanished, states wrote their own rules. Illinois amended its Human Rights Act through Public Act 103-0804. It took effect January 1, 2026.

Employers must now tell applicants when AI is used. Workers can sue directly. Penalties reach $70,000 per violation for repeat offenders (Illinois General Assembly, HB 3773).

California's own AI hiring rules took effect October 1, 2025. They extend liability to the AI vendor, not just the employer (Mayer Brown, August 2025).

Texas and Colorado passed similar laws. Each uses a different legal test for liability (National Law Review, March 2026).

$70,000

Maximum civil penalty per violation under Illinois' AI hiring disclosure law for repeat offenders (Public Act 103-0804, effective January 1, 2026).

If you're applying for a US job in Illinois, California, Texas, or Colorado, check your state law too. Everywhere else, the federal ADA is still your baseline protection, deleted guidance page or not.

Courts are still enforcing this, guidance or not

The Workday lawsuit shows the risk didn't go away. A federal court refused to throw the case out in July 2024. It let a group of rejected applicants join the case in May 2025 (Mobley v. Workday, N.D. Cal.; National Law Review, March 2026).

The case includes disability discrimination claims. It isn't just about age. The EEOC even filed a brief supporting the plaintiff, back in April 2024, before its own guidance came down.

State-level enforcement is patchy too. New York already requires bias audits of AI hiring tools under Local Law 144. A city agency received only two complaints in the law's first two years, yet auditors still found 17 likely violations just by checking employer websites (New York State Comptroller audit, December 2025).

A missing guidance page doesn't mean a missing case, but enforcement everywhere is thin. Document what happened to you: dates, screenshots, and the exact wording of any rejection.

If you're applying for a UK job instead

UK employers face a separate duty. It doesn't depend on US federal policy at all. The Equality Act 2010 requires reasonable adjustments during recruitment, including in automated selection (Equality Act 2010, sections 20-21).

That duty hasn't moved, whatever happened at the EEOC. If a UK employer's online test hurts your chances because of dyslexia, you can ask for an adjustment. The same law applies once you're hired, too.

Don't assume a US policy change affects a UK job application. UK recruiters' Equality Act duties haven't changed.

What to do if it's already happened

Save the job posting straight away. Save any description of the tool used. Save the exact wording of your rejection.

Ask the employer directly: was AI or automated screening part of this decision?

In the US, file an EEOC charge within 180 days. Some states allow 300. Our breakdown of what filing an EEOC charge actually does covers the five stages.

In Illinois, California, Texas, or Colorado, you may also have a state complaint route. The deadline is often different there.

Asking for help with a hiring test still counts as an accommodation request, even after the fact. The reasonable adjustments builder turns your specific problem into a written request. Keep a copy on file.

The paperwork trail matters more with the federal explainer gone. Write down what happened while you still remember the details.

If an online hiring test ever felt impossible to finish in time, don't wait to see if you're rejected. Ask for an accessible version in writing, first. That single email can be the only evidence you have if the employer refuses.