Before You Let AI Screen a Single Resume

The governance questions to answer first, the disclosures you owe candidates, and where automated screening quietly creates liability.

AI IN HR | 9-MINUTE READ | C-SOLUTIONS HR CONSULTING

Most employers I talk to in South Florida believe this subject doesn’t apply to them. Florida has no law governing artificial intelligence in hiring. No bias audit requirement, no candidate notice requirement, no registration, nothing. On its face, a Miami employer can run every resume through a screening tool tomorrow morning and answer to no one.

That reading is wrong, and it has nothing to do with Florida.

Your applicants bring the law with them

Employment law in this area follows the applicant and the employee, not the employer’s headquarters. If you post a remote role and someone in Illinois applies, Illinois law is in the conversation. If you have an employee working from Colorado, Colorado is in the conversation. If a New York City resident applies for a position you will fill, New York City is in the conversation.

Here is what's in force as of late 2026.

New York City Local Law 144 has been in effect since July 2023. Any automated employment decision tool used for hiring or promotion must undergo an independent bias audit each year. The audit summary must be posted publicly. Candidates must be notified at least ten business days before the tool is used and, upon request, offered an alternative selection process. Critically, the law applies even when a human makes the final decision based on the tool’s scores or rankings. Penalties run $500 to $1,500 per violation, counted per day and per affected applicant.

California Civil Rights Council regulations, effective 1 October 2025. Using an automated decision system that discriminates on a protected trait is unlawful. Employers must provide meaningful human oversight, with someone empowered to override the system, and must proactively test for bias rather than waiting to be asked. Employers must keep records for at least four years. If a system could disadvantage someone on a protected basis, employers must provide an accommodation or alternative assessment. A system that elicits disability information may constitute an unlawful medical inquiry.

Illinois House Bill 3773, effective 1 January 2026. Employers may not use AI in a way that produces bias against a protected class, whether or not the bias was intended. Employers must notify candidates and employees when they use AI in employment decisions. Using ZIP code as a proxy for a protected characteristic is expressly banned. No private right of action exists, so complaints go to the Illinois Department of Human Rights.

Colorado Artificial Intelligence Act takes effect on 30 June 2026. High-risk systems that make or influence significant employment decisions must disclose their use, conduct an annual impact assessment, and meet transparency requirements. Adverse decisions require documentation and an appeal process. Violations are treated as unfair trade practices.

Texas Responsible Artificial Intelligence Governance Act, effective 1 January 2026. Narrower than the others. It bans intentional discrimination through AI but does not allow disparate impact alone as a basis for liability. Enforcement sits exclusively with the Attorney General, and employers get a sixty-day cure period. Penalties still reach $200,000 for an uncurable violation.

A Florida employer with a remote workforce is not governed by Florida law alone. It is governed by the law of every state it hires into.

The federal picture got quieter, and that is not the same as safer

In 2023, the Equal Employment Opportunity Commission published technical assistance on artificial intelligence and Title VII. The agency has since taken that document down while it reviews its materials. If you are reading commentary written before this year, it may still point you to guidance that is no longer on the EEOC website.

What has not changed is the law underneath it. Title VII, the Americans with Disabilities Act, and the Age Discrimination in Employment Act all still apply to decisions made with a tool, exactly as they apply to decisions made by a person. A screening system that filters out applicants over forty creates age-discrimination exposure whether or not a federal agency published a fact sheet about it that month.

Guidance is a convenience. The statute is the obligation. Removing the first doesn't change the second.

Buying the tool does not transfer the risk

This is the most expensive misunderstanding I encounter. An employer buys a screening product, the vendor says it is tested and fair, and the employer treats that as the end of the matter.

It is not. Both the employer and the technology provider can be held responsible when an AI-driven decision discriminates. The vendor may carry its own liability, particularly where it effectively controls the decision, but that liability is additional to yours, not a substitute. You made the hiring decision. You are the employer.

Vendor assurance matters only if you can produce it in writing, including what it tested and when. "They told us it was fine" is not a record.

Eight questions to answer before the tool touches a resume

• What exactly does the tool do? Scoring, ranking, filtering, and keyword matching all create different exposure. Write down which one it is

• Who can override it, and have they ever? Human oversight that exists only on an org chart is not oversight

• What did the vendor test for bias, when, and against which population? Get it in writing and keep it

• What are you telling candidates, and when? Notice requirements vary, but the ten business day standard in New York City is a workable floor to build to

• What is your alternative process for someone who asks not to be screened by the tool, or who needs an accommodation?

• Does any part of the tool collect or infer health, disability, or age information? If so, you may be making a medical inquiry without realizing it

• Which states are your applicants actually in? Not where your office is. Where they are

• Where are the records, and how long are you keeping them? Four years is the California floor and a sensible default everywhere

What good governance looks like in practice

None of this requires a legal department. It requires a written policy, a named owner, a file, and the discipline to keep the file current.

The policy states which tools are approved, what they may and may not be used for, who reviews their output, and how a candidate asks for a human review. The file holds the vendor documentation, the bias testing you were given, the notice language you actually sent, the dates you sent it, and a record of every time a human overrode the system. That is the entire infrastructure.

An employer who has that file answers an agency inquiry in an afternoon. An employer who doesn’t spends weeks reconstructing what happened from memory and email, the same position you end up in after any employment claim where the documentation was never built.

The tool is not the risk. The absence of a record showing how you used it is the risk.

Where to start

If you are already screening with AI and none of the above exists, start with an inventory. List every tool touching a hiring decision, what it does, who owns it, and which states your applicants come from. That list will tell you in about an hour whether you have a small documentation project or a real problem.

If you have not started using these tools yet, you are in a better position. Build the policy before the tool, not after it. It costs almost nothing now, and it is expensive to retrofit once decisions have already been made.

A practical next step

C-Solutions publishes an AI Use Policy and Governance Pack for employers who want the documentation in place without building it from nothing. It contains six working documents:

1. AI use policy template for employees

2. Candidate and employee disclosure language

3. Vendor due diligence question set

4. Bias and adverse impact review checklist

5. Manager guidance on AI-assisted decisions

They are editable Word documents, not locked PDFs, so you can adapt them to the states you actually hire in. $99 at csolutionshrconsulting.com/resources.

If you would rather talk it through first, the site offers a free fifteen-minute consultation. You will leave it knowing whether you have an exposure worth addressing.

C-Solutions HR Consulting provides human resources consulting and is not a law firm. This article is general information about a rapidly changing area of law, current as of October 2026, and is not legal advice. Requirements differ by jurisdiction and change frequently. Before relying on any tool or policy described here, have qualified employment counsel review it against the laws of every state in which you hire.

C-Solutions HR Consulting | csolutionshrconsulting.com | Cleona@csolutionshrconsulting.com