When a computer plays a real part in deciding whether you get a job, you ought to be told, you ought to be able to fix a mistake, and you ought to be able to ask a person to take a second look. That's the whole idea behind the Alabama Workforce Algorithmic Fairness and Transparency Act — basic fairness in AI hiring that a small business can live with.
Read the plan ↓If an employer uses an automated tool as a substantial factor in a hiring, promotion, or termination decision, the plan asks for a short list of common-sense protections:
Workers are told when an automated tool is used and, if rejected, given the main reasons.
Workers may correct inaccurate data and request review by a person.
Employers under 15 employees are exempt; those with 15 to 24 face only the notice-and-appeal rules.
An independent third-party audit applies only to the largest employers (100 or more); mid-sized employers (25 to 99) face a lighter duty — a written policy and a periodic in-house check for bias. All phased in over two years.
No criminal penalty; trade secrets protected; at-will employment and right-to-work expressly preserved; a 30-day notice-and-cure window before any private lawsuit, with damages capped.
None. This is about transparency, not putting business owners in jeopardy.
Employers keep full control over their decisions. The plan governs the process — whether you're told and whether you can respond — not the outcome.
At-will employment and Alabama's right-to-work guarantees are expressly preserved. Nothing here changes them.
Employers and software vendors don't have to hand over proprietary code or confidential information.
No new bureaucracy, and the smallest employers carry essentially none of it. The duties scale up only with the size of the company.
When you apply for a job in Alabama today, there's a decent chance the first thing to read your application isn't a person. It's a program.
Employers — from national retailers to school districts to mid-sized companies right here in the Birmingham area — increasingly use artificial intelligence to screen resumes, rank candidates, score personality questionnaires, and even analyze recorded video interviews for things like word choice and facial movement. Used well, these tools save time and can help busy hiring managers look at more people, not fewer.
But there's a quieter side to this story, and it's the reason I'm releasing a policy plan today.
The risk with automated hiring tools isn't that the robots are coming for us. It's much more ordinary than that: a machine can make or heavily influence a decision about your livelihood, get it wrong, and leave you with no idea it even happened.
This isn't hypothetical. One of the largest companies in the world spent years building an AI recruiting tool, only to discover it had taught itself to downgrade applications that included the word “women's” — as in “women's chess club captain.” The company scrapped the project in 2018. The point isn't that the company was evil. The point is that even a sophisticated, well-resourced team built a tool that quietly discriminated, and it took serious effort to catch it. The tool was learning from past hiring data, and past hiring data carried old biases forward.
Now imagine a smaller employer using an off-the-shelf screening tool it bought from a vendor. It has neither the engineers nor the budget to audit that tool. It may not even fully understand how it ranks people. And the applicant who gets filtered out? She usually never learns a machine was involved at all.
Here's the gap that should bother all of us, regardless of party: federal law already prohibits discrimination in the outcome of a hiring decision. But nothing in federal law requires an employer to tell you that an automated tool was used, to explain the decision, to let you correct inaccurate data, or to give you a human second look. You can be screened out by a system that made a factual error about you — a wrong graduation date, a misread resume — and never get the chance to say, “Wait, that's not right.”
That's not a partisan observation. It's just a hole in the floor.
The proposal I'm releasing is called the Alabama Workforce Algorithmic Fairness and Transparency Act. The name is a mouthful; the idea is short.
If an employer uses an automated tool as a substantial factor in a covered employment decision — hiring, promotion, discipline, termination — the plan would require four plain things:
That's the heart of it. Notice, explanation, correction, appeal. A person, not just a program.
Just as important is what the plan deliberately does not do:
I want to be very direct about the concern I expect to hear first: isn't this just more regulation that raises the cost of hiring?
I built this plan specifically to answer that, because I run small businesses myself and I know what compliance costs feel like.
In other words, the burden is tiered to match the size of the employer, and the smallest businesses carry essentially none of it.
There's also a built-in safe harbor. If an employer follows a recognized national risk-management standard — specifically the Artificial Intelligence Risk Management Framework from the National Institute of Standards and Technology — they get a legal presumption that they acted reasonably. And here's the Alabama part: that's the exact framework Alabama's own Generative Artificial Intelligence Task Force, created by executive order, recommended in its 2025 final report. This isn't an outside standard being imposed on us. It's the standard Alabama already chose for itself.
A fair question: hasn't another state already done this? Yes — and that's where the most useful lesson lives.
In 2024, Colorado passed the first comprehensive state AI law. It was broad, covering many industries at once, and it landed with a heavy compliance burden. The business community pushed back hard. And before the law ever took effect, Colorado went back and repealed and replaced it with a lighter, narrower version.
We don't have to relive that. We can look at where Colorado started, look at where it ended up, and start at the smarter destination. My plan is deliberately narrower than Colorado's original — it covers employment only, not eight different sectors — and it's calibrated to the lighter touch Colorado itself ultimately chose. Alabama can skip the costly first draft and go straight to the version that actually works for both workers and employers.
The plan can be enforced two ways. The Attorney General can act under Alabama's existing Deceptive Trade Practices Act — the same consumer-protection law already on the books — with civil penalties on Alabama's existing scale. And an individual who's been wronged can bring a private claim.
I know the private-claim piece will draw scrutiny, so I built guardrails right into it: there's a 30-day notice-and-cure window — meaning an employer who fixes the problem when notified faces no lawsuit at all — damages are capped, and a first violation can't be turned into a class action. This mirrors protections Alabama's consumer-protection law already provides. The goal is to give the rule real teeth without opening the door to opportunistic litigation.
I'll be straight about my own stake here. I build AI software, and I'm a teacher — and AI screening has worked its way into education hiring too, including for teaching applicants. So this comes from lived experience on both sides: someone who builds these tools and someone who's watched them sort real people.
That's exactly why I don't approach this from fear. I'm not anti-technology; I use this technology every day. The question is never “is AI good or bad.” The honest answer is “both can be true at once” — these tools genuinely help, and they can genuinely harm, sometimes in the same afternoon. The job of good policy is to keep the upside and add a backstop for the downside.
This is a draft, and I mean that. Before anything like this would ever be introduced, I want to hear from Alabama workers, from civil-rights and disability advocates who see the sharp edges first, and from the business owners who'll have to live with whatever we write. The full text is below. Read it. Push on it. Tell me where I got it wrong.
A person — not just a program — ought to have a say in your livelihood. Let's build the Alabama version of that, together.
This is the complete working draft of the Alabama Workforce Algorithmic Fairness and Transparency Act, published here so any Alabamian can read exactly what is proposed. It is a preliminary draft for public discussion — see the note at the bottom.
A BILL
TO BE ENTITLED
AN ACT
Existing law provides no state-law protections governing how artificial intelligence and automated systems are used to make employment decisions about Alabama workers and job applicants. Federal anti-discrimination laws apply to the outcome of a hiring decision but do not require an employer to tell an applicant that an automated system was used, to explain an adverse decision, to let the applicant correct inaccurate data, or to provide a human review.
This bill would establish the Alabama Workforce Algorithmic Fairness and Transparency Act. It would require an employer that uses an automated employment decision tool as a substantial factor in a covered employment decision to: notify the applicant or employee that such a tool was used; if the decision is adverse, provide the principal reasons and an opportunity to correct inaccurate personal data and to appeal for review by a human; and, for larger employers, maintain a written policy and periodically test the tool for discriminatory effect. This bill would phase in these requirements by employer size, would exempt the smallest employers, would provide safe harbors for employers that follow nationally recognized risk-management standards and that test for and correct bias, and would provide for enforcement by the Attorney General under the Alabama Deceptive Trade Practices Act and by a private right of action with notice-and-cure safeguards. This bill would not create any criminal penalty.
Relating to employment and to consumer protection; to add a new Chapter to Title 25, Code of Alabama 1975 (subject to placement by the Code Commissioner), to be known as the Alabama Workforce Algorithmic Fairness and Transparency Act; to define terms relating to artificial intelligence and automated employment decision tools; to require notice to an applicant or employee when such a tool is a substantial factor in a covered employment decision; to require, upon an adverse decision, a statement of reasons and an opportunity to correct data and to appeal for human review; to require larger employers to maintain a written governance policy and to test covered tools for discriminatory effect; to provide safe harbors and exemptions, including for small employers; to provide for enforcement by the Attorney General and by a private right of action; to provide that the act creates no criminal penalty; and to provide for phased effective dates.
BE IT ENACTED BY THE LEGISLATURE OF ALABAMA:
This act shall be known and may be cited as the “Alabama Workforce Algorithmic Fairness and Transparency Act.”
The Legislature finds and declares the following:
(a) Alabama employers increasingly use artificial intelligence and automated systems to screen applications, rank candidates, analyze video interviews, and recommend hiring, promotion, and termination decisions. These tools can expand opportunity and reduce paperwork, but when they operate without transparency they can also screen out qualified Alabamians for reasons no person ever reviews.
(b) Documented cases — including a major retailer's recruiting tool that was found to downgrade applications associated with women and was abandoned in 2018 — show that automated hiring tools can reproduce or amplify bias present in their training data.
(c) Alabama school districts, small businesses, and large employers alike now rely on third-party screening tools. Teachers, hourly workers, and professionals deserve to know when a machine, rather than a person, played a substantial part in a decision about their livelihood, and to have a fair chance to correct an error and ask a human to take a second look.
(d) Alabama has already recognized the importance of responsible artificial intelligence. The State's Generative Artificial Intelligence Task Force, established by Executive Order 738 (2024), recommended in its 2025 final report that artificial intelligence be governed by transparency and by the National Institute of Standards and Technology Artificial Intelligence Risk Management Framework. This act extends that same principle of responsible, transparent use to the employment context.
(e) This act is intended to protect Alabama workers and applicants while respecting employers, particularly small businesses, by tailoring obligations to employer size, by recognizing nationally accepted risk-management standards as a safe harbor, by protecting trade secrets and proprietary information, and by imposing no criminal penalty. This act does not alter the at-will employment doctrine or the right-to-work guarantees of Sections 25-7-30 to 25-7-35A, Code of Alabama 1975, and governs only the transparency and fairness of the process by which covered decisions are made.
For purposes of this act, the following terms have the following meanings:
(1) ADVERSE EMPLOYMENT DECISION. A covered employment decision that results in the denial to an applicant or employee of, or a material limitation on, employment or an employment opportunity, including a decision not to hire, not to advance, or to terminate.
(2) ARTIFICIAL INTELLIGENCE SYSTEM. A machine-based system that, for any explicit or implicit objective, infers from the inputs it receives how to generate outputs, including content, decisions, predictions, or recommendations, that can influence physical or virtual environments.
(3) AUTOMATED EMPLOYMENT DECISION TOOL. An artificial intelligence system that is used to make, or is a substantial factor in making, a covered employment decision. The term does not include a tool described in subsection (5) of Section 9.
(4) COVERED EMPLOYER. A person doing business in this state that employs 15 or more individuals in this state and that uses an automated employment decision tool with respect to an individual in this state. Employer tiers are defined in Section 6.
(5) COVERED EMPLOYMENT DECISION. A decision that has a material effect on the provision or denial to an individual of, or the terms of, employment or an employment opportunity, including hiring, promotion, compensation level, reassignment, discipline, or termination.
(6) DEVELOPER. A person doing business in this state that develops, or intentionally and substantially modifies, an automated employment decision tool and offers, sells, licenses, or otherwise makes it available to an employer.
(7) DISCRIMINATORY EFFECT. A condition in which the use of an automated employment decision tool results in unlawful differential treatment of, or unlawful disparate impact on, an individual or group on the basis of a characteristic protected under federal law or under the laws of this state, including age as protected by the Alabama Age Discrimination in Employment Act, Sections 25-1-20 to 25-1-29, Code of Alabama 1975. The term does not include either of the following:
a. An employer's good-faith self-testing conducted to identify, prevent, or mitigate discrimination or to ensure compliance with law.
b. An action taken to expand an applicant or employee pool to increase diversity or to redress historical discrimination.
(8) SUBSTANTIAL FACTOR. A factor generated by an artificial intelligence system that assists in making a covered employment decision and is capable of altering the outcome of that decision. The term includes any content, score, ranking, prediction, or recommendation generated by an artificial intelligence system concerning an individual that is used as a basis for a covered employment decision.
(9) TRADE SECRET. Has the meaning given in the Alabama Trade Secrets Act, Section 8-27-2, Code of Alabama 1975.
(a) Before, or at the time, a covered employer uses an automated employment decision tool as a substantial factor in a covered employment decision concerning an individual, the employer shall notify the individual, in plain language, of all of the following:
(1) That an automated employment decision tool will be or was used as a substantial factor in the decision.
(2) The type of decision and a plain-language description of the general type of the tool and the general categories of personal characteristics or data it assesses.
(3) Contact information for the employer and instructions for exercising the rights provided in Section 5.
(b) The notice required by this section shall be provided directly to the individual, in plain language, in any language in which the employer customarily recruits for the position, and in a format accessible to individuals with disabilities.
(a) If a covered employment decision in which an automated employment decision tool was a substantial factor is adverse to the individual, the covered employer shall, within a reasonable time and not later than 30 days after the decision, provide the individual all of the following:
(1) A statement of the principal reason or reasons for the decision, including the degree to which and manner in which the automated employment decision tool contributed to the decision and the general type and source of the data the tool processed concerning the individual.
(2) A reasonable opportunity to correct any inaccurate personal data that the tool processed in making, or as a substantial factor in making, the decision.
(3) A reasonable opportunity to appeal the decision and to obtain review by a natural person, if such review is technically feasible. Human review need not be provided where it is not in the best interest of the individual, including where a delay would pose a risk to the life or safety of any person.
(b) An employer may satisfy subdivision (a)(1) by providing a clear, standardized, plain-language notice. Nothing in this section requires disclosure of a trade secret or of information protected from disclosure by state or federal law; if the employer withholds information on that basis, it shall so inform the individual and state the general basis for the withholding.
In addition to Sections 4 and 5, which apply to all covered employers, the following obligations apply according to employer size:
(a) Small covered employer (15 to 24 employees in this state). No obligation beyond Sections 4 and 5.
(b) Medium covered employer (25 to 99 employees in this state). The employer shall maintain a brief written policy, reasonable in light of the employer's size, describing the automated employment decision tools it uses and the steps it takes to identify and mitigate discriminatory effect, and shall review each covered tool for discriminatory effect at least once every 24 months. The employer shall make a summary of the policy available to applicants and employees on request.
(c) Large covered employer (100 or more employees in this state). The employer shall comply with subsection (b) on an annual basis and, in addition, shall obtain an evaluation of each covered tool for discriminatory effect by an independent third party not less than once every 24 months. The independent evaluation requirement does not apply to a tool for which the developer has, within the preceding 24 months, made available to the employer a substantially equivalent independent evaluation.
A developer that offers, sells, licenses, or otherwise makes available an automated employment decision tool to a covered employer in this state shall make available to the employer the information reasonably necessary for the employer to comply with this act, including a plain-language statement of the tool's purpose and intended uses, the general categories of data used to develop the tool, any known limitations or known risks of discriminatory effect, and a description of how the tool was evaluated for discriminatory effect. A developer is not required to disclose a trade secret or information protected from disclosure by law.
(a) Standards-based presumption. In any action under this act, there is a rebuttable presumption that a covered employer or developer used reasonable care to protect individuals from discriminatory effect if the employer or developer substantially complied with this act and with the most current version of the Artificial Intelligence Risk Management Framework published by the National Institute of Standards and Technology, or another nationally recognized risk-management framework that is substantially equivalent or more protective.
(b) Discover-and-cure affirmative defense. It is an affirmative defense to an action under this act that the covered employer or developer discovered and cured the alleged violation as a result of feedback, adversarial testing or red teaming as those terms are used by the National Institute of Standards and Technology, or an internal review process, and is otherwise in substantial compliance with a framework described in subsection (a). The party asserting the defense bears the burden of demonstrating that its requirements are satisfied.
(c) Self-testing protected. Good-faith self-testing and pool-expansion described in Section 3(7)a. and b. do not constitute a violation of this act and may not be used as evidence of a violation.
This act does not apply to, and the term “automated employment decision tool” does not include, any of the following:
(1) An employer that employs fewer than 15 individuals in this state.
(2) An artificial intelligence system intended only to perform a narrow procedural task, or to detect decision-making patterns or deviations from prior patterns, that is not intended to replace or influence a previously completed human assessment without sufficient human review.
(3) Anti-fraud technology that does not use facial recognition, anti-malware, antivirus, calculators, cybersecurity tools, databases, data storage, firewalls, networking, spam- and robocall-filtering, spell-checking, spreadsheets, web hosting, and similar technologies, unless the technology is a substantial factor in making a covered employment decision.
(4) Activity to the extent necessary to comply with federal, state, or local law; to respond to a governmental or law-enforcement inquiry, subpoena, or summons; to investigate, establish, exercise, prepare for, or defend a legal claim; to protect life or physical safety; or to prevent, detect, or respond to security incidents, fraud, or illegal activity by means other than facial recognition technology.
(5) An automated employment decision tool used by or for the federal government, or developed, approved, or certified by a federal agency acting within its authority, except where the tool is a substantial factor in a covered employment decision.
(6) Nothing in this act shall be construed to impose any obligation that adversely affects rights guaranteed by the First Amendment to the United States Constitution or by Section 4 of Article I of the Constitution of Alabama of 2022.
A person claiming an exemption under this section bears the burden of demonstrating that the action qualifies for the exemption.
(a) Attorney General. A violation of this act constitutes an unlawful trade practice for purposes of the Alabama Deceptive Trade Practices Act, Sections 8-19-1 to 8-19-15, Code of Alabama 1975, and may be enforced by the Attorney General as provided in that chapter, including by injunctive relief and by a civil penalty not to exceed two thousand dollars ($2,000) for each violation as provided in Section 8-19-11, Code of Alabama 1975.
(b) Private right of action. An applicant or employee aggrieved by a violation of Section 4 or Section 5 may bring a civil action, but only after providing the employer written notice of the specific alleged violation and a period of not less than 30 days to cure it. If the employer cures the violation within that period and provides the individual the rights to which the individual was entitled, no action may be maintained for that violation. In an action under this subsection, a prevailing individual may recover actual damages or five hundred dollars ($500) per violation, whichever is greater, together with reasonable attorney's fees and costs.
(c) Limitations on private actions. A first violation by an employer may not be the basis for a class action. The affirmative defense and presumption in Section 8 apply to an action under this subsection.
(d) No criminal penalty. This act does not create, and may not be construed to create, any criminal penalty.
(e) No effect on other rights. This act does not preempt or limit any other right, claim, or remedy available under law, including under federal anti-discrimination law or the Alabama Age Discrimination in Employment Act, Sections 25-1-20 to 25-1-29, Code of Alabama 1975.
The Attorney General may adopt rules under the Alabama Administrative Procedure Act, Chapter 22 of Title 41, Code of Alabama 1975, solely to prescribe the form and manner of the notices and statements required by Sections 4 and 5. This act does not authorize rulemaking that expands the substantive duties established by this act.
(a) Sections 4 and 5 (notice, explanation, correction, and human appeal) become effective 12 months after the date this act is enacted, and apply to all covered employers.
(b) Section 6(b) (medium-employer policy and periodic testing) and Section 7 (developer disclosure) become effective 18 months after the date this act is enacted.
(c) Section 6(c) (large-employer annual testing and independent third-party evaluation) becomes effective 24 months after the date this act is enacted.
(d) Except as otherwise provided, this act becomes effective on the first day of the third month following its passage and approval by the Governor, or its otherwise becoming law.
It's built to avoid exactly that. Employers under 15 employees are fully exempt. Those with 15 to 24 just have to tell the applicant a machine was involved and let them appeal to a person — nothing more. Audit duties only reach the largest employers (100 or more), and even those phase in over two years.
Colorado tried the heavy version, business pushed back, and they walked it back. This plan starts where they landed.
It doesn't. The plan governs process transparency, not who you hire. At-will employment and right-to-work are expressly preserved.
If a machine plays a substantial role in the decision: tell the person, let them fix an error, and let them ask for a human review. That's it. The hiring decision stays entirely with the employer.
Federal law covers the outcome of discrimination, but it requires no notice, no explanation, no data correction, and no human review. You can be screened out by a system that got a basic fact about you wrong and never even know a machine was involved.
That's the specific gap this plan fills — and nothing more.
Both can be true at once. These tools cut paperwork and can expand opportunity, and they can quietly screen people out with no human watching — sometimes in the same afternoon.
This plan doesn't take a side on whether AI is good or bad. It keeps the upside and adds a backstop for the downside.
Studied it as a model, deliberately didn't copy it. This plan is employment-only (Colorado's original covered eight sectors), written in Alabama legal style, and tied to Alabama's own NIST task-force standard. It's lighter on small business and calibrated to the lighter version Colorado itself ended up adopting.
I want to hear from Alabama workers, from civil-rights and disability advocates, and from the business owners who'd have to live with whatever we write. Read it, push on it, and send your thoughts — the good and the critical.
Share your feedback →