What Illinois's AI Rule Changes—and What It Still Leaves Open
Public Act 103-0804 is best understood as a change to the Illinois Human Rights Act's way of seeing employment automation. The central question is not whether Illinois has declared every use of artificial intelligence unlawful, nor whether a disclosure can make an automated process fair. It is what follows when an employer uses AI in a covered employment setting and the law separates the result of that use, the role of a named geographic input, and the information given to an employee. This final synthesis draws those threads together so the reader can make a high-level judgment that is precise about enacted duties and equally precise about what still depends on rules, facts, and interpretation.
The amendment creates a three-part accountability model
The strongest supported conclusion is that Public Act 103-0804 creates a three-part legal architecture inside the Illinois Human Rights Act. Section 2-102(L)(1) addresses an employer's use of artificial intelligence that has the effect of subjecting employees to discrimination on the basis of protected classes under the Article. The same paragraph separately names the use of zip codes as a proxy for protected classes. Section 2-102(L)(2) makes failure to provide notice to an employee that the employer is using artificial intelligence for the paragraph (1) purposes a separate civil-rights violation.
These are related duties, but they are not interchangeable. They attach to different features of an employment process.
The effects question concerns the consequence of the employer's AI-assisted practice. Did a ranking, prediction, recommendation, generated evaluation, or other output influence recruitment, hiring, promotion, training selection, discharge, discipline, tenure, or a term, privilege, or condition of employment in a way that subjects employees to discrimination on a protected basis? The phrase “has the effect” makes consequence central. The provision does not state an additional requirement that the employer intended to discriminate. But the text also does not supply the complete proof method: it does not set a disparity threshold, comparison group, sample-size rule, causation standard, or approved metric.
The proxy question concerns the role of zip code in the process. Was the zip code used as a stand-in for a protected class? That question follows the input path and the employer's use of the variable. It is not merely a second name for an outcome statistic. A process may have an effects concern without using zip code. It may use zip code in a way that raises a proxy question before the available data can establish an outcome difference. It may raise both questions. The statute names the proxy pathway so that it does not disappear when reviewers look only at final selections.
The notice question concerns information. Did the employer provide notice to an employee that it was using AI for the purposes described in paragraph (1)? Notice can make an otherwise invisible feature of the process visible, but it does not answer what the process did or how zip code entered it. The statute does not say that notice cures a discriminatory effect or a proxy use. It places transparency alongside substantive accountability.
This separation is more than a drafting detail. A single automated hiring process can contain all three kinds of facts. Imagine an employer uses a system to rank applicants, the system receives zip codes, the ranking influences who receives an interview, and the employer does not disclose the AI use. The employer's purpose may be efficiency, and a recruiter may make the final decision. None of those facts collapses the inquiry. A reviewer must still look at the outcome, trace the geographic variable, and identify the notice event. Each record answers a different question.
The architecture also changes what “neutral” can mean. A neutral business objective describes why an employer adopted a tool. A uniform formula describes how the tool is applied. Neither description establishes the effect on protected groups, explains whether zip code was used as a proxy, or proves that the required notice was given.