Are Pre-Employment Assessments Legal? What Contractors Need to Know
This article is general information for contractors. It is not legal advice. Employment law varies by state and city and changes often, so check your own situation with an employment attorney.
The short answer: yes. Pre-employment assessments are legal in the United States, and federal agencies have recognized their use since the 1970s. The obligations sit with you as the employer: use the assessment the same way for every candidate for a seat, keep it related to the job, avoid anything that works as a medical examination before an offer, provide accommodations when asked, watch your results for adverse impact, and follow any state or city rules on automated hiring tools where you hire.
That last item has changed the most. New York City, Illinois, California, Colorado, Connecticut and New Jersey now each have rules in force or on the calendar that reach at least some hiring tools, and several of them reach scored tools whether or not AI is involved. A contractor hiring in one of those places has more to do than a contractor hiring in most of the country.
The federal baseline
Title VII and adverse impact
Title VII of the Civil Rights Act prohibits employment discrimination based on race, color, religion, sex and national origin. It covers employers with 15 or more employees. It reaches practices that look neutral but screen out a protected group at a higher rate. The Supreme Court established that principle in Griggs v. Duke Power Co. (1971), a case about hiring tests and diploma requirements.
The practical tool for spotting this is the four-fifths rule from the federal Uniform Guidelines on Employee Selection Procedures. If one group's selection rate is less than 80 percent of the rate for the group with the highest rate, federal agencies will generally treat that as evidence of adverse impact.
A simple example: if 10 of 20 candidates from one group pass a screen (50 percent) and 6 of 20 from another group pass (30 percent), the ratio is 30 divided by 50, or 60 percent. That falls below four-fifths and deserves a closer look.
Adverse impact starts a second question: whether the tool is job-related and consistent with business necessity, and whether an equally useful alternative with less impact was available.
Federal agencies have pulled back from adverse impact enforcement since 2025. That changes less for an employer than it sounds. Candidates can still bring adverse impact claims under Title VII in private lawsuits, the Uniform Guidelines and the four-fifths rule are still on the books, and some states now write adverse impact into their own rules. A federal court in California has also let claims go forward against a hiring software vendor as the employer's agent (Mobley v. Workday), so the same questions are reaching the tools themselves.
The ADA
The Americans with Disabilities Act prohibits medical examinations and disability-related questions before a job offer. An assessment that diagnoses a mental health condition or asks about one can cross that line. Clinical personality instruments are the usual concern.
The ADA also requires reasonable accommodation in the hiring process. If a candidate asks for more time, a different format or another adjustment because of a disability, you need a process to handle it.
Timed assessments need a specific plan. Decide ahead of time what you will offer a candidate who asks for more time, such as an extended or untimed sitting, and confirm your tool can deliver it. California's rules on automated decision systems name tools that measure reaction time as ones that may call for accommodation. When a request comes in, record what was asked and what you provided, and keep any medical information you receive in a separate confidential file, apart from the hiring file, as the ADA requires.
The ADEA
The Age Discrimination in Employment Act protects applicants 40 and older, and it covers employers with 20 or more employees. The same logic applies: consistent use, job-related content, and attention to whether older candidates are screened out at higher rates.
Records
Federal rules generally require employers to keep hiring records, including test results used in hiring, for at least one year. Some states require longer.
California's rules for automated decision systems call for four years. If a charge of discrimination is filed, federal rules require you to keep the related records until it is resolved. Keep more than the score: the notice the candidate received and the date it went out, the report, any accommodation request and what you provided, and the reason for your decision. If a hiring tool lets you delete a candidate, export what you need first, because deleting from the tool may remove the only copy you have.
State and city rules on automated hiring tools
These laws were written mostly with AI resume screeners and video interview analysis in mind. Their definitions vary, and some are broad enough to reach scoring tools that use statistics or formulas. Whether a particular tool is covered depends on the definition and how you use it. Ask your attorney.
New York City, Local Law 144 (enforced since July 2023). Covers "automated employment decision tools" used to substantially assist hiring or promotion decisions for jobs in the city. Employers must have a bias audit completed within one year before use, publish a summary of the results, and notify candidates at least 10 business days before the tool is used.
Illinois, HB 3773 (effective January 1, 2026). Amends the Illinois Human Rights Act to prohibit using AI in employment decisions in a way that discriminates, and requires employers to notify applicants and employees when AI is used in covered decisions. The Illinois Department of Human Rights proposed detailed notice rules and then pulled them back for more work, so the law is in effect while the final rules are still pending. Watch for them.
California (effective October 1, 2025). Regulations under the state's Fair Employment and Housing Act address automated decision systems in employment. The definition reaches systems built on algorithms and statistics as well as AI. Employers must keep the related data for four years, anti-bias testing (or the lack of it) can count as evidence in a discrimination claim, and a vendor acting for the employer can be treated as the employer's agent.
Colorado (effective January 1, 2027). Colorado replaced its 2024 AI Act with a new law in May 2026. Its definition covers technology that uses computation on personal data to produce a score used in a decision like hiring, so a tool can be covered with no AI in it. As enacted, it requires employers using covered automated decision tools in hiring to give advance notice, explain an adverse decision in plain language within 30 days, offer a way to correct inaccurate data and a path to human review, and keep records for three years. It also puts duties on the companies that sell these tools, including written documentation for employers on the tool's limits and how human review should work. The Attorney General is expected to issue rules before it takes effect. A federal court blocked the earlier Colorado law in April 2026 and a challenge to this one is expected, so plan as if it starts on time and watch the date.
Connecticut (employment notice duties from October 1, 2027). Public Act 26-15, signed in May 2026, uses a similar definition built on computation. Employers using a covered tool in hiring must give candidates written notice before it is used, including the product's trade name. Sellers of these tools must give employers the information they need to comply.
New Jersey (in force since December 15, 2025). Rules under the state's Law Against Discrimination spell out disparate impact liability, including for tests and automated tools used in hiring. A practice that screens out a protected group at a higher rate can be unlawful with no intent to discriminate, unless it serves a substantial, legitimate, nondiscriminatory interest and no less discriminatory alternative would meet that interest.
If you hire in any of these places, or anywhere with a similar rule, get a specific answer from counsel before you roll out any scored hiring tool.
A practical checklist for contractors
- Same seat, same tool, same process. Every candidate for a given opening gets the same assessment, the same questions and the same steps. Consistency is your best protection and it makes candidates easier to compare.
- Keep it job-related. Choose tools built around the seat you are filling, and be able to explain in plain English why each step relates to the work.
- Use it as one input. Make the decision yourself, using the assessment alongside the interview, working interview and references. Write down why you made the decision.
- Offer accommodations. Tell candidates they can ask for one, and have a named person who handles requests.
- Tell candidates what to expect. What the assessment is, how long it takes, who sees the results, and how their data is handled. In some places this is required.
- Watch your numbers. Once you have enough candidates for a seat, compare pass and hire rates across groups using the four-fifths rule.
- Keep records. Store assessment results and hiring decisions for at least the federal minimum, and longer where your state requires.
- Ask vendors direct questions. What the tool measures, whether AI is involved in scoring, what evidence supports it, whether a bias audit or adverse impact study exists, how long candidate data is kept, and what documentation they provide for state rules on automated hiring tools.
- Keep a copy of what you told candidates. Save the notice each candidate received and the date it went out. Where notice is required, that copy is your proof.
- Export before you delete. Before removing a candidate from any hiring tool, save the report and the answer record for as long as your longest record rule requires.
How First Move Assessment handles these questions
First Move Assessment publishes its answers on the site in plain English.
- It is a pre-hire assessment of first instinct for a specific seat. It is not a medical or psychological examination, a personality inventory, a skills exam or a background check.
- Scoring runs on fixed arithmetic. No machine learning model produces the score, and no candidate data goes to an AI provider.
- No bias audit or adverse impact study has been conducted, and the product makes no EEOC compliance claim.
- You make every hiring decision. The report gives you a seat match with its margin of error, the places the candidate pulls away from the seat, and interview questions for those places.
- An untimed version is available at no extra cost when a candidate asks.
- Completed assessments are kept for four years or until you delete them, and deleting a candidate removes their name, email, answers and results.
- Whether your use triggers any state or local obligation is your responsibility to determine, and the site says so.
Each trade page, from HVAC and plumbing to commercial cleaning, carries the same limits in plain English and links to the full legal section.
Frequently asked questions
Are pre-employment assessments legal?
Yes. They are legal under federal law when used consistently, related to the job, and monitored for adverse impact. Some states and cities add notice, audit or recordkeeping rules for automated hiring tools.
What is the four-fifths rule in hiring?
A guideline from federal selection procedures: if one group's selection rate is less than 80 percent of the highest group's rate, agencies generally treat that as evidence of adverse impact, which then requires the employer to show the practice is job-related.
Can I use a personality test before making a job offer?
Many non-clinical personality inventories are used pre-offer. Instruments designed to identify mental health conditions can count as medical examinations under the ADA, which are prohibited before an offer. Check with counsel before using any clinical instrument.
Do AI hiring laws apply to small contractors?
Some do. NYC's law applies to employers using covered tools for jobs in the city regardless of size, and Illinois's applies to employers covered by its Human Rights Act, which includes employers with one or more employees. Check the definitions for your state and city.
Do I have to tell candidates I use an assessment?
In some places, yes, and it is good practice everywhere. Tell candidates what the assessment is, how long it takes, and how to ask for an accommodation.
Do state automated-hiring laws cover tools that use no AI?
Some do. California's rules reach systems built on algorithms and statistics, and Colorado's and Connecticut's new laws reach technology that uses computation to produce a score. New York City's law turns on whether a computer helps set a tool's inputs or their weights, and Illinois's covers AI. Whether a particular tool is covered depends on the definition and how you use it, so ask counsel.
How long should I keep pre-employment assessment results?
At least one year under federal rules, longer where your state requires, and until any discrimination charge is resolved. California's rules for automated decision systems call for four years. If you hire in several states, keeping hiring records for four years covers the longest rule named in this article.
What should I do when a candidate asks for more time?
Treat it as an accommodation request. Send it to the person who handles requests, offer the extended or untimed format your tool supports, and write down what was asked and what you provided. Keep any medical information in a separate confidential file.
What does the AI do?
Nothing. No AI is used when your candidate takes the assessment or when the report is made. Their answers are scored by a fixed formula. Every number that formula uses was set in advance and approved before use, and nothing in it is learned or adjusted by a computer from anybody's answers. Every written section of the report is a fixed template filled in from the numbers the formula produced. The same answers give the same result every time, and no candidate information is sent to any AI provider.
Why we left AI out. First, you can check it: every number in the report comes from arithmetic that returns the same answer every time it runs, and an AI model can give two different answers to the same question. Second, it does not drift: an AI model can be changed by its maker without notice, which would quietly change what your reports say, while this formula changes only when we publish a new version, and every report records the version that scored it. Third, your candidate's answers stay with us and are never handed to an AI company. Fourth, the law treats AI in hiring as a special case, and most of the cost of that lands on the employer who uses it.
What the law says. These are the main rules in force or already passed as of September 2026. It is general information, not legal advice.
New York City, Local Law 144 of 2021 (NYC Administrative Code section 20-870 and following; rules at 6 RCNY section 5-300). An employer may not use an automated employment decision tool unless it has had an independent bias audit within the past year, a summary of that audit is published, and candidates are given notice at least 10 business days before it is used. The rules cover tools built with machine learning, statistical modeling, data analytics or AI in which a computer at least in part picks the inputs or how much each one counts, in order to make the result more accurate.
Illinois, Human Rights Act as amended by Public Act 103-0804 (775 ILCS 5/2-102, in force since January 1, 2026). An employer must tell applicants and employees when it uses artificial intelligence in hiring and other employment decisions, may not use AI in a way that discriminates against a protected class, and may not use ZIP codes as a stand-in for one.
California, Civil Rights Council regulations on automated-decision systems (California Code of Regulations, title 2, in force since October 1, 2025). These cover any computational process that makes or helps make an employment decision, including one built on algorithms or statistics rather than AI. A tool that screens people out on a protected characteristic can be unlawful whatever it is built on, and employers must keep the related records for four years.
Colorado, Senate Bill 26-189 (in force from January 1, 2027, replacing the state's 2024 AI Act). It covers automated decision-making technology, meaning any technology that uses computation on personal data to produce scores or other output used in a decision about a person, with or without AI. Employers must give notice at the point of use, explain the tool's role within 30 days of an adverse decision, and offer human review and reconsideration on request.
Federal law. Title VII of the Civil Rights Act, the Americans with Disabilities Act and the Age Discrimination in Employment Act apply to every selection procedure, AI or not, and the Uniform Guidelines on Employee Selection Procedures (29 CFR Part 1607) set out how adverse impact is judged.
What that means for you. Leaving AI out keeps this tool clear of the parts of these laws written around AI and machine learning. It does not take it outside all of them. California and Colorado define the tools they cover by computation rather than by AI, so a fixed formula like this one can still fall within them, and federal law applies to every assessment. What applies to you depends on where you hire and how you use the report. Section 7 of the legal terms and disclaimers sets out your responsibilities, and your own employment lawyer should confirm what applies where you hire.
See what the report includes
Read three complete First Move Assessment reports and judge for yourself what a seat-based report gives you. The full legal and disclosure section is on the home page. Get the three sample reports.