What HB 3773 Changes
HB 3773 amends the Illinois Human Rights Act to address employer use of artificial intelligence, including generative AI, in employment decisions. Signed by Governor JB Pritzker in August 2024, the law takes effect January 1, 2026. It does not create a standalone AI statute; instead, it folds AI-related discrimination into the existing IHRA enforcement framework, meaning violations are pursued through the same complaint and investigation process IDHR already uses for other employment discrimination claims.
The law applies to AI used across the employment lifecycle, from recruitment and hiring through promotion, training selection, discipline, and termination. Its definition of artificial intelligence is broad enough to cover both traditional statistical or algorithmic scoring tools and generative AI systems, which expands scope beyond the narrower automated employment decision tool definitions used in some other state and local laws.
Core Obligations
Three obligations sit at the center of HB 3773:
- Non-discrimination: employers may not use AI in a way that subjects employees or applicants to discrimination based on protected classes recognized under the IHRA.
- ZIP code restriction: the law specifically prohibits using ZIP code as a proxy for protected class status when AI informs an employment decision. This implies that employers need visibility into which input features their AI tools rely on, since ZIP code and similar geographic variables can correlate with race, national origin, or other protected characteristics even when not explicitly used as inputs.
- Employee notice: employers must notify employees when AI is used to make decisions affecting them. The statute does not specify exact notice language or delivery mechanics; organizations should expect further detail through IDHR rulemaking rather than the bill text itself.
What the Law Does Not Prescribe
Unlike NYC Local Law 144, which mandates independent bias audits for automated employment decision tools, HB 3773 does not specify a bias-audit methodology or a technical testing standard. It also does not prescribe a required audit trail or logging architecture in the statute text.
The compliance obligations as written center on discrimination outcomes and notice, not on a defined system architecture. This gives employers latitude in how they implement compliance, but it also means organizations cannot point to a checklist of technical requirements in the statute itself. IDHR rulemaking was pending as of the most recent available research, and procedural or documentation requirements may be added once implementing regulations are finalized. Compliance programs built now should be treated as provisional and revisited when those rules are published.
Where Runtime Controls Fit
HB 3773 does not mandate a specific technical architecture, but its non-discrimination and notice obligations create practical dependencies on system-level visibility. Employers need to know which AI systems are actually making or influencing employment decisions at the point of use, not just what a vendor's documentation claims.
This is where runtime governance and monitoring become relevant as an operational control, distinct from legal compliance itself. Runtime policy enforcement can restrict which AI agents or tools are permitted to access employment-related data or invoke decision-influencing functions, based on least-privilege permissions rather than static role assignments. Audit logging at the runtime layer can record when an AI system was invoked, what inputs it processed, and what output it produced for a given employment action, which supports the kind of decision-level record keeping HB 3773's notice provisions imply.
Trussed AI provides runtime governance and security for enterprise AI agents, including access controls, tool approval workflows, and audit logging, which organizations can apply to AI systems operating in employment workflows as one component of a broader HB 3773 compliance program. This is not a substitute for legal review or IDHR rule compliance, but it addresses the technical visibility gap that statute text alone does not resolve.