Sponsored by AIntel Compliance
Find out which of your AI tools these laws already reach.
The free AI Compliance Readiness Scorecard takes two minutes. Ten questions, a score out of 20, and every gap ranked from most to least urgent, including whether your organization tells people when AI helps decide who gets hired, explains a no in plain language and offers a human review.
Take the free scorecard →AIntel Compliance is a sister brand of The RN Network, so no commission changes hands this week — the sponsor slot is house-run.
The short version
- 57% of HR professionals in states with workforce AI laws don’t know those laws exist. Of those who do, 12% have acted.
- Recruiting is where HR’s AI use is most concentrated, which puts the nurse requisition at the front of the exposure.
- Colorado’s new law takes effect January 1, 2027 and covers any tool that scores, ranks, recommends or classifies a candidate.
- A tool that only schedules, routes or summarizes is excluded. One automatic knockout question can move it across that line.
- A nurse you reject can ask for a plain-language explanation within 30 days, a correction and a human review.
Somewhere inside your applicant tracking system there is a setting that ranks nurses. Maybe it is a match score beside every name. Maybe it is a knockout question that quietly drops anyone without two years of med-surg. Maybe it is the chatbot your team switched on last spring to book interviews after hours.
Your recruiters know exactly where it is. Your compliance team almost certainly does not.
That gap stops being a paperwork problem on January 1, 2027.
Most HR teams working under an AI hiring law do not know the law exists
SHRM’s State of AI in HR 2026, drawn from more than 1,900 HR professionals, asked people working in states with workforce-related AI regulations whether they knew the rules. 57% did not. Of the ones who did, 12% had taken steps to comply.
Recruiting is not on the sidelines of that finding. SHRM puts recruiting at 27%, the largest of the three areas where HR’s AI use is concentrated, ahead of HR technology at 21% and learning and development at 17%. The same research counts 19 of the most populous states with AI laws or regulations that touch employment.
Put those findings side by side and the likeliest place for a health system to be out of compliance without knowing it is the nurse requisition.
Three jurisdictions already regulate AI in hiring, and Colorado joins them on January 1
New York City has required a bias audit of automated employment decision tools, and notice to candidates, since 2023. Illinois amended its Human Rights Act effective January 1, 2026, so employers must give notice whenever AI influences recruitment, hiring, promotion or discharge, and may not use zip codes as a stand-in for protected characteristics. The state’s draft rules would put that notice in the job posting itself, name the product and its vendor, and keep the record for four years.
California’s civil rights regulations brought automated-decision systems under the Fair Employment and Housing Act on October 1, 2025, with a four-year record requirement of their own.
Health systems that hire across state lines, and the staffing firms that recruit for them, are often inside more than one of these at once. A single remote case management opening can be read by a nurse in Chicago, a nurse in Denver and a nurse in Los Angeles on the same afternoon.
Colorado’s law turns on whether a tool routes candidates or ranks them
Colorado signed SB 26-189 on May 14, 2026, replacing its 2024 AI Act with a narrower law that takes effect January 1, 2027. It reaches automated decision-making technology that materially influences a consequential decision, and employment decisions are on the list. A tool is covered when it produces a score, a ranking, a recommendation, a classification or a prediction used to make or assist that decision. Résumé scoring and interview recommendations are the examples practitioners keep naming.
What sits outside is just as specific. Tools that only summarize, organize, translate, draft, route or present information for a human are excluded, as long as they produce no score or ranking that materially influences the outcome.
Here is why that distinction matters to a nurse recruiter. In January 2019 Houston Methodist piloted a recruiting chatbot named Mia for RN roles. It reported a chat-to-offer-accept time of 27.9 days for RNs against an RN hiring norm of 80 days, and 52% of candidates chatted with it outside normal business hours. Those are seven-year-old pilot figures, and they still make the case for conversational recruiting better than most vendor decks do. Time to fill is still the number your hiring leaders are judged on, and our time-to-fill benchmarks in the Workforce Intelligence Center carry it by specialty and region if you want your own market beside that pilot.
My read: a chatbot that answers questions and books interviews looks like routing under Colorado’s definition. Add one automatic knockout question that ends the conversation for a nurse without a compact license, and the same tool is producing the classification that decides who moves forward.
Colorado’s attorney general has until January 1, 2027 to define “materially influence” by rule, so that line has not been drawn yet. Your vendor’s sales deck has not drawn it either.
A nurse you reject can ask you three things, and the clock on the first one is 30 days
Before a covered tool is used, Colorado expects clear and conspicuous notice that it is in play and how to learn more. A prominent notice at the point of interaction may be enough.
After an adverse decision, the employer owes a disclosure within 30 days. It has to describe the decision and the tool’s role in plain language, offer a simple way to ask about the tool and the inputs it used, and explain the person’s rights. The candidate can then ask to correct personal data that is factually wrong, and can ask for human review by someone with the authority to approve, change or override the outcome, trained for the job and not simply deferring to the tool’s output.
Records that show compliance have to be kept for at least three years after the decision, including the tool’s version identifiers and change logs. Enforcement belongs to the attorney general alone, with no private right of action and a 60-day cure period for curable violations until 2030. Existing discrimination claims, disparate impact included, stay exactly where they were.
On a high-volume RN requisition, the nurses you turn away outnumber the ones you hire many times over. Every one of them is a potential 30-day clock if a covered tool touched the decision.
Staffing firms carry this exposure for every client they recruit for
Illinois’ draft rules reach employers and anyone acting for them, recruiters and agents included. Colorado goes a step further on contracts. Liability between the company that built a tool and the one using it is split by relative fault, and indemnification clauses that shift it are invalid, except against a user who ran the tool in a way its developer never intended.
If your firm screens nurses with its own tools and hands a hospital a ranked shortlist, settle who the deployer is before January, and put the answer in the contract.
Six moves a talent team can make before January 1
Inventory every tool that touches a nurse application. Applicant tracking match scores, knockout questions, chatbots, assessments, video interview platforms and sourcing tools that rank profiles. Mark each one “routes” or “ranks.”
Get the answer from each vendor in writing. Ask whether the tool produces a score, ranking, recommendation or classification, which version you are running, and whether a change log exists. You will need version identifiers in your records for three years.
Put the notice where the nurse applies. Illinois expects it in the posting itself, and Colorado accepts a prominent notice at the point of interaction. Writing it once for the posting and the apply page is the simplest place to start.
Write the 30-day letter now. The decision in plain language, the tool’s role, and how to ask for more. Decide who sends it before the first rejection goes out in January.
Name the human reviewer. Pick someone trained who can override the outcome and will look at the evidence rather than re-read the score. On most nurse requisitions that is a recruiting lead or TA manager, not a hiring manager who never saw the tool.
Audit every knockout question. Years of experience, shift availability and license type are the usual ones. Each becomes a classification the moment it ends an application automatically, so treat each one as covered until the attorney general says otherwise.
The nurses you turn away are about to be able to ask why
For as long as I have been on the hiring side of the table, the rejected candidate has been the one person in the funnel nobody had to answer to. A form email went out, or nothing did, and the requisition moved on.
Starting in January, a nurse in Colorado who never got past your screen can ask what decided it, and someone on your team has to know the answer. Find out this month which of your tools would be in that answer.