The Illinois Department of Human Rights announced on June 2, 2026, that the public hearing scheduled for June 10 regarding proposed Subpart J regulations under HB 3773 has been "temporarily postponed" pending "continued collaboration with other state agencies." The IDHR provided no rescheduled date. The proposed rules — which would require employers to give written notice to employees and their bargaining representatives whenever AI influences a covered employment decision — remain in a 45-day written comment period that closes June 29, 2026. HB 3773 has been operative since January 1, 2026, and prohibits the discriminatory use of AI in hiring, promotion, discipline, RTW determinations, and the terms and conditions of employment. The postponement introduces regulatory uncertainty, but attorneys representing injured workers whose employers or carriers deploy AI-driven RTW or claims tools should not treat this as enforcement relief — the statute itself is already in effect.
Reporter's Take: The phrase "continued collaboration with other state agencies" is doing a lot of work here. It suggests IDHR may be aligning with Illinois DOL or the Governor's office on scope questions — possibly as to whether AI-driven workers' comp claims triage falls within the employment decision definition. That question matters enormously for WC practitioners, and the answer is likely yes.
Every time a legal AI agent starts a new task, it starts without memory of the prior 10,000 tasks your firm has run through the same platform. That's the "context tax" — the cost of re-establishing institutional knowledge at the start of every session. It's why an AI that drafts a brilliant contract clause today may draft a subtly inconsistent one tomorrow, because it has no access to what your firm has approved, argued, or rejected in past matters.
Harvey's acquisition of DeepJudge in early 2026 was explicitly targeted at this problem. DeepJudge had built a knowledge-graph architecture for institutional legal memory — a way to make a firm's prior work product, playbooks, and approved positions continuously available to AI agents without having to re-inject them into every new context window. Harvey's Command Center interface now surfaces this institutional knowledge layer alongside its 25,000+ custom agents, so an agent drafting a coverage position letter can access how your firm has argued that same coverage issue across 400 prior cases.
For WC/PI firms, the near-term practical version of this is simpler: the firms that are building structured case outcome databases now — documenting what worked, what didn't, what settlement ranges by injury type, which arbitrators respond to which framing — are building the raw material for institutional memory AI. The firms treating every closed case as a transaction to be filed and forgotten are not.
The Fifth District affirmed in part, reversed in part, and remanded in this wage-differential credit dispute. The core issue: whether an employer is entitled to a credit for scheduled-loss payments (§8(e)) when a claimant subsequently recovers under §8(d)(1) wage differential. The court held that where a claimant with a line-of-work injury recovers under §8(d)(1), the employer is entitled to the credit. But under the IWCC's reading of §8(d)(2), that credit does not apply — an asymmetry the court found textually defensible but logically strained.
The June 10 public hearing has been postponed. Written comments remain open through June 29. Key provisions: (1) written notice required before any AI-influenced employment decision; (2) notice must go to the employee's bargaining representative; (3) third-party vendors acting on the employer's behalf are treated as the employer for notice purposes. The IDHR has not indicated whether it will revise the draft rules before finalizing.
The original Colorado AI Act (SB 24-205) has been stayed by federal court and effectively replaced by SB 26-189. The replacement narrows compliance obligations to a transparency framework, eliminates private lawsuits, and delays effective date to January 1, 2027. Employers with Colorado workers who were preparing June 30 compliance packages should pivot entirely. Dozens of additional state AI bills remain active nationally.
The IDHR's June 10 hearing postponement has caused some employer-side practitioners to treat the Subpart J rulemaking as paused. It is not. HB 3773 has been in effect since January 1, 2026. The proposed regulations interpret and implement a statute that is already operative. An employer using AI to evaluate return-to-work fitness, score job applications from injured workers, or run claims-severity triage today has potential IHRA exposure regardless of when final rules issue.
Illinois WC attorneys representing injured workers should be asking two questions on every new matter: (1) Did the employer or carrier use AI in any decision affecting the claimant's employment terms, RTW assessment, or claim handling? (2) If so, did the claimant receive any written notice of that AI use? The absence of notice is, under HB 3773's plain text, a violation — and the proposed rules make clear that carriers and their vendors acting on an employer's behalf are within scope.
AI-enhanced MRI and 3D reconstruction technology is expanding what IME physicians can detect in a single examination — including subtle disc degeneration, ligament micro-tears, and soft-tissue findings that previously required specialist radiologist consultation. But the litigation risk runs in both directions. Correlation is not causation, and a sophisticated AI imaging tool that identifies pre-existing degenerative changes may be used to argue that a claimant's current condition is unrelated to the workplace injury — even where the degenerative findings were clinically silent before the accident.
Plaintiff counsel should be preparing to challenge AI imaging conclusions on Daubert/Frye grounds, including requests for the underlying algorithm's training data, validation methodology, and accuracy benchmarks. Defense counsel and carriers using AI-assisted IME QA should be aware that the audit trail generated by those QA tools — including any flagged inconsistencies in the original report — may be discoverable.
The Thomson Reuters Institute's 2026 report documents that organizations nearly doubled their use of generative AI in the past year, with more than half of professionals now relying on publicly available tools. The report includes legal-sector specific data on where AI investment is concentrating — and a detailed look at the Thomson Reuters / Supio partnership as a model for specialty-area AI integration in PI litigation.