All the intelligence that's fit to argue.
The nation's most-watched AI employment law died before it could live. Colorado Governor Jared Polis signed SB 26-189 on May 14, 2026 — a wholesale replacement of the state's landmark 2024 Colorado AI Act, which had been scheduled to take effect June 30. The original law would have imposed bias audits, full risk management programs, and algorithmic discrimination assessments on any employer deploying AI in hiring, promotion, or termination decisions. After sustained pushback from the tech industry, a constitutional challenge filed by xAI (joined by the DOJ), and a federal magistrate's April 27 stay, the Colorado legislature passed the rewrite in the final days of its session. The replacement law strips the substantive obligations down to transparency-only: employers must now merely disclose that AI is being used in covered decisions, with no mandate for bias testing or impact assessment. Illinois, meanwhile, is moving in the opposite direction: IDHR's Subpart J comment window on HB 3773 implementation rules closed June 29, and the June 10 public hearing was postponed June 2 for "continued collaboration with other state agencies" — a delay that signals substantive rule revision is underway, not retreat.
Reporter's Take: Colorado's retreat from substantive AI accountability creates a federal regulatory vacuum that Illinois is uniquely positioned to fill — and the state's BIPA enforcement history suggests it will. For WC/PI practitioners whose carrier and TPA clients use AI in claims triage, RTW evaluation, and disability management, the Illinois trajectory is the one to watch. The June 29 comment deadline is not an abstraction; it's the last opportunity to shape rules that will govern your clients' exposure.
The personal injury legal market reached $61.7 billion in 2026 with 60% of firms now using AI tools operationally — a majority threshold that signals the end of the experimental phase and the beginning of baseline-expectation status for AI in PI practice, per the LawPro.ai/Morgan & Morgan Future of Legal Tech 2026 report. Why it matters: Firms not yet operationally deploying AI are now below the market median, not ahead of it. [ECIKS]
Lavern, developed by Innanen AI, formally released its open-source agentic law firm platform under Apache 2.0 license — deploying 67 specialized AI agents covering intake through settlement — marking the first significant open-source agentic legal framework available for self-deployment by practices and legal AI builders. Why it matters: Open-source architecture lets WC/PI shops self-host or customize agent workflows without SaaS vendor dependency; the 18-month signal timeline matters for planning.
The 2024 study cited in this week's 9th Circuit hallucination opinion found Westlaw AI hallucinating 17% of answers and Lexis AI hallucinating 33% — rates that apply to the most widely used legal research tools in practice. Why it matters: These are not fringe tools; WC/PI practitioners relying on AI-assisted research in any of these platforms must build independent verification workflows into every filing. [Bloomberg Law]
IDHR postponed its public hearing on HB 3773 / Subpart J AI employment rules June 2, citing the need for "continued collaboration with other state agencies." The 45-day comment window triggered by May 15 Illinois Register publication remains open through June 29. Why it matters: For WC practitioners, RTW and employment decision AI tools used by carriers and employers are squarely in scope; now is the time to review and comment. [IDHR]
The Fifth District Appellate Court set aside an IWCC award in WC Subway v. IWCC (April 13, 2026), remanding on grounds that the commission's finding of work-relatedness was not supported by the manifest weight of evidence — a recurring causation sufficiency challenge that defense WC counsel continue to successfully deploy on appellate review. Why it matters: Causation documentation at the arbitration level remains the linchpin; AI-assisted medical chronologies that surface gaps or inconsistencies are now standard defense preparation tools.
LawPro.ai's new practitioner guide argues PI firms are at the same inflection point the profession hit when offices moved from typewriters to PCs — and the window to build competitive advantage is closing quarterly. The 56% of firms spending under $5K/year on AI are deferring costs that will compound.
AI-enhanced MRI imaging is now appearing in IME reports as a clinical tool, producing findings that weren't visible on standard reads. Both plaintiff and defense counsel should be asking IME physicians to disclose which imaging platforms and AI-enhancement tools were used — and whether those outputs are discoverable.
Thomson Reuters' 2026 AI in Professional Services Report confirms generative AI use nearly doubled in the past year, with over half of professionals now relying on publicly available tools. The enterprise credibility signal: TR's direct Supio partnership validates the WC/PI-specific AI category to institutional buyers.
What it does: LawPro.ai is a PI-focused AI platform anchored by its TimeLineD™ technology, which extracts causation timelines from medical records and builds structured chronologies, and its Case Assistant Casey, which handles intake, document review, and case strategy support. The platform includes Hallucination Prevention Technology™ (HPT), which grounds AI outputs in verified source documentation before presenting results to users.
WC/PI use case: Extract a causation timeline from a 400-page medical record in under 60 seconds, with every finding source-cited to the original document. Casey then helps prepare case-strategy analysis and demand narrative from the same record set.
Verdict: The HPT architecture directly addresses the verification problem the 9th Circuit just made a disciplinary issue — if it performs as described, it's a meaningful differentiator in a market full of tools that don't account for output accuracy. → lawpro.ai
Lavern's formal open-source launch under Apache 2.0 is the most significant agentic legal architecture release to date. The platform deploys 67 specialized agents covering the full case lifecycle — intake triage, medical record organization, discovery management, deposition preparation, settlement valuation, and demand letter generation — orchestrated through a central case management agent that routes tasks to specialist sub-agents based on case context.
The Apache 2.0 license means any practice or vendor can fork, deploy, and customize the framework without licensing fees. For WC/PI-specific builders, this is a foundation to build on: connect a Lavern intake agent to an IWCC e-filing integration, add a causation-analysis sub-agent trained on Illinois WC caselaw, and the result is a jurisdiction-specific agentic system at a fraction of custom-build cost.
The 18-month signal: firms that begin experimenting with agentic deployment now will have working integrations while competitors are still evaluating vendor RFPs. The window is open; it won't stay open indefinitely.
Fifth District sets aside IWCC benefits award and remands on manifest weight causation grounds. Commission's work-relatedness finding held insufficient; case returned for re-evaluation of causation evidence. Confirms the appellate court's continued willingness to reverse commission findings where medical evidence does not clearly support work-relatedness.
Significance: Defense WC counsel should continue to build strong causation evidentiary records at arbitration, as appellate reversal on manifest weight grounds remains a viable strategy when commission awards are poorly supported.
IDHR postponed June 10 public hearing on proposed Subpart J rules implementing HB 3773's AI employment notice and non-discrimination requirements. Comment window through June 29 remains open. Law in effect since January 1, 2026; enforcement exposure for non-compliant employers ongoing while rules are finalized.
Significance: RTW decision AI tools used by carriers and self-insured employers may be in scope; assessment of AI tool inventory recommended before comment deadline.
Colorado replaced the 2024 Colorado AI Act (which would have required bias audits and risk assessments) with a transparency-only framework requiring only disclosure of AI use in covered employment decisions. The original law was stayed by federal court on April 27 following constitutional challenges from xAI and DOJ. The rewrite eliminates substantive obligations and takes effect immediately.
Significance: The Colorado retreat is a data point in the federal-vs-state AI regulation tension; employers with national carrier clients should monitor whether other states follow or diverge.
Illinois HB 3773 / Public Act 103-0804 has been in effect since January 1, 2026. The law prohibits the use of AI that has the effect of discriminating against employees in covered employment decisions — including RTW determinations, discipline, and the "terms, privileges, or conditions of employment" — even if the discrimination is unintentional. It also prohibits using zip codes as proxy for protected classes.
For WC practitioners, the critical question is whether AI tools used by employers, carriers, and TPAs in RTW and accommodation decisions are covered. The IDHR's proposed Subpart J rules (pending final publication after the June 10 hearing postponement) define "make or influence covered employment decisions" broadly. A carrier-side AI tool that evaluates an injured worker's RTW fitness and feeds that evaluation to an employer's HR system is almost certainly in scope.
Violations expose employers to actual damages, civil penalties, attorneys' fees, and compliance reporting obligations — the same remedial framework as other IHRA civil rights violations. With comment window open through June 29, this is the moment for WC stakeholders to engage.
AI-enhanced MRI imaging is now appearing in IME reports as a clinical adjunct — platforms that apply machine learning to standard imaging to surface findings not visible on conventional reads. For plaintiff counsel, this creates a new discovery obligation: when an IME physician uses an AI imaging tool to support a causation opinion, the platform, its training data, its validation studies, and its output in this specific case are potentially discoverable and Daubert-challengeable.
Defense counsel using these tools to bolster denial opinions face the same exposure in reverse. The evidentiary landscape is shifting: AI-enhanced imaging findings may be powerful, but they require the same foundational expert disclosure as any other methodology. "The IME physician used an AI imaging assistant" is not a complete disclosure.
Begin asking in IME depositions: which imaging AI platform was used, was it FDA-cleared for the claimed use, what is its published sensitivity and specificity, and what did it flag in this case that the standard read did not?
A practitioner's guide to competing in the AI-transformed PI market, written by a working PI attorney who co-founded LawPro.ai. More candid than typical vendor content — the author argues the window for building AI competitive advantage is closing quarter by quarter and names the structural shifts (PE consolidation, AI-native firm formation) that are compressing it. Worth 20 minutes regardless of whether you use LawPro's tools. → PRNewswire announcement & download link
When the most ambitious state AI employment law in the country collapses before its effective date, it creates a standard-setting vacuum. Illinois is the most likely state to fill it — not because of political coincidence but because of institutional capacity: BIPA enforcement infrastructure, a functioning IDHR rulemaking process, and a legislative history of moving first on technology employment regulation. WC practitioners whose carrier clients operate nationally need to calibrate their compliance exposure to Illinois, not Colorado.
17% and 33% hallucination rates for the two dominant legal research platforms, cited by the Ninth Circuit in a published discipline order, means roughly 1 in 6 to 1 in 3 AI-generated legal research answers contain errors. That is not a marginal quality issue; it's a structural workflow risk. The answer is not to stop using AI research tools — it's to build verification into the workflow as a non-negotiable step, the same way you cite-check before filing. The attorneys sanctioned this week failed at that step. Don't.
More than half of PI firms spend under $5,000 annually on AI. That's less than one associate's monthly salary, less than a single expert witness deposition, less than many court reporter bills. The LawPro.ai data doesn't suggest these firms aren't aware of AI — 92% say they're using it. They're using it at the lowest possible cost tier: generic tools, no integration, no workflow investment. The gap between that tier and firms building real operational AI capability is widening every quarter. The median is not safe.
The Illinois Workers' Compensation Act's permanent partial disability schedules assign specific numbers of weeks to each body part lost or impaired — the thumb is worth 76 weeks, the entire arm 253 weeks, and loss of hearing in one ear 54 weeks. These scheduled values are fixed regardless of the claimant's wage or occupation and cannot be combined with a wage differential award under Section 8(d)(1) per Azcon Metals v. IWCC.
Agentic AI (n.) — An AI system that autonomously performs multi-step tasks by planning, selecting tools, and taking sequential actions toward a defined goal, without requiring a human to direct each step. Distinguished from chatbots that respond to single prompts. Lavern's 67-agent architecture is a legal example: a central orchestration agent routes tasks to specialist agents (intake, discovery, settlement) that operate independently within their domain.
"It's a mistake to believe that any one player can do it alone, even a frontier lab."
— Jason Boehmig, incoming OpenAI product lead for the legal vertical, on joining OpenAI to build "Codex for Legal," June 1, 2026