Illinois HB 3773 amends the Human Rights Act to regulate AI in employment decisions. Disparate impact prevention, notice obligations, consent tracking, and audit trails mapped to SWT3 witness procedures.
Who this is for: HR technology vendors deploying AI hiring tools in Illinois, employment law counsel advising on AI compliance, chief people officers and HR directors at companies with Illinois employees, and third-party auditors evaluating AI employment system fairness.
Enforcement: January 1, 2026 (in effect). HB 3773 amends the Illinois Human Rights Act. Applies to any employer with 1 or more employees in Illinois during 20 or more calendar weeks. Covers all employment decisions: recruitment, hiring, promotion, discharge, discipline, tenure, and terms of employment. The 2026 amendment requires explicit written consent for AI video interview analysis (replacing implicit consent). Enforced by the Illinois Department of Human Rights.
Illinois HB 3773 amends the Illinois Human Rights Act to explicitly regulate the use of artificial intelligence in employment decisions. The law prohibits employers from using AI in a manner that produces a discriminatory effect on applicants or employees based on protected class membership.
HB 3773 covers every stage of the employment lifecycle where AI is used. Unlike narrower laws that target only hiring (e.g., NYC LL 144), Illinois applies the prohibition across all employment decisions.
| Employment Decision | Covered | Example |
|---|---|---|
| Recruitment | Yes | AI resume screening, job ad targeting |
| Hiring | Yes | AI interview scoring, candidate ranking |
| Promotion | Yes | AI performance evaluation for advancement |
| Discharge | Yes | AI-driven termination recommendations |
| Discipline | Yes | AI behavioral monitoring and flagging |
| Tenure | Yes | AI retention risk scoring |
| Terms/conditions | Yes | AI-driven compensation, scheduling, benefits |
Each HB 3773 obligation maps to one or more SWT3 witness procedures. Anchors generated by these procedures create the cryptographic evidence trail that demonstrates compliance.
| HB 3773 Obligation | SWT3 Procedure | What It Witnesses |
|---|---|---|
| No disparate impact | AI-FAIR.1 | Bias measurement: disparity percentage across protected classes |
| Protected class parity | AI-FAIR.2 | Fairness score against minimum threshold |
| Demographic balance | AI-FAIR.3 | Outcome distribution verification |
| Notice to employees | AI-TRANS.1 | Transparency disclosure evidence with timestamp |
| Explanation of AI use | AI-EXPL.1 | Explanation generation for each decision |
| Human review | AI-HITL.1 | Human review completion for employment decisions |
| Override logging | AI-HITL.2 | When human overrides AI recommendation |
| Video consent | AI-CONSENT.1 | Explicit consent collection and verification |
| Data destruction | AI-DATA.4 | Video data deletion within 30 days of request |
| Audit trail | AI-AUDIT.1 | Complete decision audit trail with anchors |
HB 3773 requires: AI must not produce a discriminatory effect on protected classes in any employment decision. The prohibition applies regardless of whether the employer intended to discriminate. Each protected class must be evaluated independently.
SWT3 witnesses: Factor A = maximum allowable disparity percentage. Factor B = measured disparity. A FAIL verdict triggers automatically when the measured disparity exceeds the threshold. Each protected class is measured independently, producing separate anchors per class per decision type.
Request AI-FAIR.1 anchors filtered by model. A model with 95%+ pass rate across 90+ days demonstrates ongoing disparate impact monitoring, not just pre-deployment testing. Look for continuous coverage across all seven employment decision types.
HB 3773 requires: Employers must notify employees before using AI for employment purposes, explaining what AI is used, what it evaluates, and how its output factors into the decision. The notice must be provided before the AI-assisted decision occurs.
SWT3 witnesses: Notice delivery with cryptographic timestamp. The anchor proves when the notice was provided relative to the AI decision. Factor B >= Factor A confirms that the notice preceded the decision. The timestamp chain is immutable and independently verifiable.
Compare AI-TRANS.1 anchor timestamps against employment decision dates. Notices must precede decisions. Post-decision notices do not satisfy HB 3773. Look for gaps where decisions lack corresponding notice anchors.
HB 3773 requires: Meaningful human review of AI employment decisions, particularly for adverse actions such as termination, discipline, or denial of promotion.
SWT3 witnesses: Human review completion. Factor A = decisions requiring review. Factor B = reviews completed. A FAIL verdict is generated when reviews are incomplete. Each review is tied to a specific employment decision via the anchor chain.
Look for AI-HITL.1 anchors correlated with adverse employment actions. Missing human review anchors before a termination or discipline decision indicates a compliance gap. Cross-reference with AI-HITL.2 to verify that overrides are also logged.
HB 3773 requires (via AIVIA amendment): Explicit written consent before AI analysis of video interviews. Continuing the interview no longer constitutes consent under the 2026 amendment. This is a significant change from the original AIVIA, which allowed implicit consent.
SWT3 witnesses: Consent collection status. Factor A = 1 (consent required). Factor B = 1 (consent obtained) or 0 (not obtained). A FAIL verdict blocks AI analysis without consent. The anchor records the consent mechanism, timestamp, and applicant identifier.
In gatekeeper mode (strict=True), the SDK blocks inference until AI-CONSENT.1 returns PASS. This enforces consent-before-analysis at the code level, preventing the AI system from processing video data without verified consent.
HB 3773 incorporates and strengthens the Artificial Intelligence Video Interview Act (AIVIA). The following rules apply specifically to AI analysis of recorded video interviews:
AI-DATA.4). The deletion must be verifiable and the evidence trail must confirm the data no longer exists.| Examiner Question | Where to Look |
|---|---|
| Is disparate impact monitored? | AI-FAIR.1 -- disparity measurements across protected classes |
| Are employees notified before AI use? | AI-TRANS.1 -- notice timestamp vs. decision timestamp |
| Is human review documented? | AI-HITL.1 -- review completion count vs. required count |
| Is video consent collected? | AI-CONSENT.1 -- consent status before AI analysis |
| Are overrides logged? | AI-HITL.2 -- human override audit trail |
| Is video data destroyed on request? | AI-DATA.4 -- deletion evidence within 30-day window |
| Is there a complete audit trail? | AI-AUDIT.1 -- decision chain with all procedure anchors |
For full SDK documentation and integration examples, see the SDK Docs. To start witnessing AI employment decisions today, create a free account.