Illinois HB 3773 Is Live, the Notice Rules Were Just Pulled, and You Still Have to Comply
Part of AI Regulation News, our running brief on the laws reshaping AI at work.
Here is the situation most Illinois employers have not fully absorbed. On January 1, 2026, HB 3773 took effect. It amended the Illinois Human Rights Act so that using artificial intelligence in a way that discriminates against a protected class in employment is now its own civil rights violation, and it added a duty to notify people when AI is used in covered employment decisions. Then, on June 2, 2026, the Illinois Department of Human Rights temporarily postponed the proposed rules that were meant to spell out exactly how the notice should look. The comment window that had been running was paused with it.
The trap is reading that second event as a reprieve. It is not. The rulemaking is in limbo, but the statute is not. The law is in force right now, and the Department of Human Rights can act on it whether or not the implementing rules are final. So you are living in the worst version of a compliance moment: a binding obligation with a deadline already behind you, and the official instruction manual sitting in a drawer. This briefing is about what HR actually owes this quarter while the rules sit unresolved.
Key takeaways
- HB 3773 has been in force since January 1, 2026. AI that discriminates in employment is now a violation of the Illinois Human Rights Act, full stop.
- The law requires you to notify employees and applicants when AI is used in covered employment decisions, and to keep records for the relevant limitation period.
- The proposed notice rules were temporarily withdrawn on June 2, 2026, but the statute still binds you. The pause delays the instructions, not the duty.
- The safe move is to comply with the plain text of the statute now, in writing, rather than wait for final rules that may look much like the draft when they return.
What HB 3773 actually requires
Strip away the rulemaking noise and the statute itself is readable. It does two things that change your job.
First, it makes AI-driven discrimination an explicit civil rights violation. If an employer uses AI in a way that has a discriminatory effect on a protected class with respect to recruitment, hiring, promotion, discipline, discharge, or the terms of employment, that is a violation of the Illinois Human Rights Act. The law also reaches the use of zip codes as a proxy for a protected class, which closes a common loophole in screening tools. You do not get to hide behind "the vendor's model did it." If the tool you chose produces a discriminatory result, the exposure is yours.
Second, it imposes a notice duty. Employers have to tell employees and applicants when AI is being used in connection with a covered employment decision. The exact form, timing, and wording of that notice is what the withdrawn rules were going to pin down. That is the piece sitting in limbo. But the duty to give notice lives in the statute, not in the rules, so the absence of final rules does not erase the obligation. It just removes the official template you were hoping to copy.
Underneath both duties is a recordkeeping reality. To defend either one, you need to be able to show, after the fact, where AI touched an employment decision and what you told people about it. That means a durable log, not a memory. Build it for the four-year horizon that civil rights claims tend to run on, and you will not be reconstructing it under a subpoena two years from now.
Live law, rules in limbo: what that means for compliance now
This is the part that confuses people, so be precise about it. A statute and its administrative rules are two different things. The statute is the law the legislature passed. The rules are the agency's detailed instructions for following it. When the Illinois Department of Human Rights postponed the proposed rules on June 2, 2026, it pressed pause on the instructions. It did nothing to the law. Effective dates do not roll back because a draft regulation got pulled.
So what does a careful HR leader do with a binding statute and no final rules? You comply with the words of the statute and you document that you did. The notice duty exists, so give a clear, written notice now rather than waiting for a form that may not arrive for months. The anti-discrimination duty exists, so test your AI-touched decisions for disparate impact now. When the rules return, they will most likely track the draft that was already circulating, which means work you do today against the statute is very unlikely to be wasted. The downside of acting early is a little rework. The downside of waiting is a live violation with your name on it.
Illinois is not an outlier here, which is why this belongs on your desk and not your watch list. Colorado's AI employment law runs in the same direction, Texas TRAIGA adds another state rulebook, and the federal posture in the EEOC enforcement plan for AI hiring tools shows the same scrutiny landing from a different angle. A multistate employer that solves Illinois cleanly has built most of what the others demand.
What HR owes now: a checklist
You can do all of this against the statute without waiting on the rules. Build it as a repeatable process, because the obligation is permanent and the next reduction or hiring sprint is already coming.
- Inventory every place AI influences or facilitates an Illinois employment decision: sourcing, screening, ranking, interview scoring, promotion, discipline, and discharge. You cannot give notice about a use you have not found.
- Draft and issue a written AI-use notice to applicants and employees now, in plain language, covering what tool is used and at what stage. Do not wait for the official template.
- Run a disparate-impact screen on every AI-touched selection step, and explicitly check for zip code or other proxies for a protected class.
- Pin vendor accountability in writing. Get the tool maker to document how its model was tested and to stand behind it, because the statute does not let you outsource the liability.
- Stand up a four-year recordkeeping log that captures where AI was used, what notice was given, and which human made the final call.
- Name a human decision-maker of record for every covered decision, so you can show a person owned the outcome, not a model output.
Guardrails
A few lines keep this from going sideways. Do not treat the withdrawn rules as permission to stop; the statute is the obligation and it is live. Do not let "the vendor handles compliance" stand as your answer, because HB 3773 attaches the violation to the employer who used the tool. Do not generate a notice once and forget it, because as your AI use spreads, a notice that was accurate in January is stale by summer. And do not let an AI assistant write a legal notice you then file without a human reading every word. Use the tool to draft fast and to keep the language consistent across locations, then have a person who owns the risk sign off before anything goes out.
This is exactly the kind of defensible, repeatable workflow you can build with an AI assistant working for you instead of against you. Claude can help you map where AI touches each decision, draft a plain-language notice that fits the statute, frame the disparate-impact questions, and keep the recordkeeping log consistent as the rules evolve. The tool moves faster and stays more consistent than a human doing it by hand. It does not decide, and it does not sign. That balance, AI doing the drafting while a named human keeps the judgment, is the entire premise of The Leveraged HR Professional, our course on building AI-assisted HR systems that hold up under exactly this kind of scrutiny.
What AI does not replace
Be clear about the line. AI can draft your notice, screen for proxies, and stand up the log. It cannot decide whether a given use of AI in hiring is lawful, it cannot be the decision-maker of record, and it cannot carry the liability if a screening tool produces a discriminatory result. HB 3773 is written to put that responsibility on the employer and the human running the process. A notice generated by software and pushed out without a person standing behind it is not compliance; it is automation with no accountability, which is precisely what the law is reaching for.
So let the assistant do the drafting, the inventory, and the consistency checks. Keep the judgment, the accountability, and the signature with a named human being. Getting good with these tools is not about handing the decision to software. It is about walking into a Department of Human Rights inquiry with a process clean enough that you would be comfortable explaining every step of it.
Not sure which AI skills your role needs first?
Take the two-minute course finder and get a path built for your work.
Join The Leverage Club for $49.
Working sessions, templates, and briefings like this one for senior professionals putting AI to work the right way.
Frequently asked questions
If the Illinois notice rules were withdrawn, do I still have to comply with HB 3773?
Yes. HB 3773 took effect January 1, 2026, and the statute remains in force. On June 2, 2026 the Illinois Department of Human Rights postponed the proposed implementing rules, which delays the detailed instructions, not the underlying law. The notice duty and the anti-discrimination duty both live in the statute, so the agency can still enforce them. The safe path is to comply with the plain text of the statute now and document that you did.
What does HB 3773 require Illinois employers to do?
It makes AI-driven discrimination in employment an explicit violation of the Illinois Human Rights Act, including the use of zip codes as a proxy for a protected class. It also requires employers to notify employees and applicants when AI is used in connection with a covered employment decision such as hiring, promotion, discipline, or discharge. To defend both duties, employers need to inventory where AI is used, give written notice, screen for disparate impact, and keep records for the relevant limitation period.
Can I blame the vendor if our AI hiring tool produces a discriminatory result?
No. HB 3773 attaches the violation to the employer that uses the tool, so "the vendor's model did it" is not a defense. You should still pin vendor accountability in writing and require documentation of how the tool was tested, but the legal exposure for a discriminatory outcome stays with you. That is why a human review gate and a disparate-impact screen on every AI-touched decision matter.
Related briefings: Colorado AI hiring law: what HR owes now · The EEOC enforcement plan for AI hiring tools · Texas TRAIGA: the AI law HR needs to read