All the intelligence that's fit to argue.
The most significant AI discipline order in federal appellate history landed Wednesday when the Ninth Circuit suspended California immigration attorneys Mike Singh Sethi and William Rounds for six months from practicing before the court, fined each $2,500, and ordered mandatory AI-use disclosure in all future filings for two years. The case, LNU v. Blanche (No. 24-4790), arose from briefs containing fabricated case citations that don't exist — including "Eduardo v. Garland, 28 F.4th 742 (9th Cir. 2022)" and "Lay v. Holder, 729 F.3d 962 (9th Cir. 2013)" — plus misattributed quotations from real opinions. What elevated the misconduct from negligence to discipline: the attorneys repeatedly denied AI involvement, called the errors "typographical mistakes," and compounded the deception at every subsequent procedural opportunity — through a motion to correct, oral argument, and response to the order to show cause. The court identified a second, harder-to-detect category of AI error — "inaccuracies" where real cases are cited but their holdings are misstated — calling these potentially more dangerous than outright fabrications because they survive cursory checks. Citing a 2024 study finding that Westlaw and Lexis AI hallucinated 17% and 33% of answers respectively, the court issued its order explicitly as a warning to the entire Ninth Circuit bar.
Reporter's Take: This opinion is doing doctrinal work. The fabrication/inaccuracy taxonomy gives courts a framework to classify and escalate AI misconduct — and the 6-month suspension for the cover-up, not just the error, sends a clear message that how you respond when AI burns you matters as much as the initial mistake. Workers' comp and PI counsel filing in federal courts should read this opinion in full and show it to every associate and paralegal touching briefs.
OpenAI formally entered the legal AI product market by hiring Jason Boehmig, co-founder and former CEO of contract management platform Ironclad ($3.2B valuation), to lead product development for a new legal vertical the company is calling "Codex for Legal." Why it matters: OpenAI joins Anthropic (Claude for Legal plug-ins) and Microsoft (Legal Agent) in building dedicated legal AI infrastructure — the platform wars have moved from foundation models to practice-specific tooling, and WC/PI-focused vendors now compete not just with each other but with trillion-dollar incumbents. [Law.com]
A Western District of Tennessee federal judge sanctioned a Memphis law firm for AI misuse arising in a malpractice action against Baker Donelson, ordering cost reimbursement and referral to state disciplinary counsel — the second AI-discipline action in a week. Why it matters: Two separate courts, two AI sanctions orders, same week. The enforcement trend is no longer episodic; it's becoming a pattern WC/PI counsel cannot dismiss as outlier events. [Law360]
The Trump DOL's February 2026 proposed rule to rescind the Biden-era 2024 independent contractor standard and revert to a two-factor economic-dependence test has moved past comment deadline (April 28) without a final rule — but state classification rules remain strict and enforcement continues. Why it matters: For WC plaintiffs' counsel, any DOL rule that tilts toward contractor status creates defense arguments against WC coverage for gig workers; monitor state-level exposure, which is often broader than federal guidance. [DOL.gov]
UK firm Mishcon de Reya announced a "bonus boost" structure taking effect July 2027 that rewards associates for hours above target and AI utilization, following Shoosmiths' precedent of tying firmwide bonuses to Copilot usage. Why it matters: Compensation structures are beginning to formalize AI productivity expectations — a directional signal for US firms, where billable-hour pressure will eventually collide with AI efficiency gains. [City AM]
The personal injury legal market hit $61.7 billion in 2026 with 60% of firms reporting AI tool use, yet the LawPro.ai/Morgan & Morgan Future of Legal Tech 2026 report shows 56% of PI firms spend $5,000 or less annually on AI — signaling a massive underinvestment gap relative to the market opportunity. Why it matters: The competitive advantage window is still open; firms willing to invest above the $5K median are building meaningful operational moats. [PRNewswire]
AI-generated demand letters are now standard at high-volume PI shops, but ABA Formal Opinion 512 requires independent verification of every fact and citation before filing. The distinction between "AI-assisted" and "AI-reviewed" is becoming a malpractice exposure line.
Carrier-side AI fraud detection platforms now cross-reference billing patterns across jurisdictions, flagging providers who consistently upcode or bill after discharge dates. Expect more aggressive pre-litigation investigations targeting provider fraud in repeat-claimant files.
The legal AI software market is now valued at $5.59 billion in 2026, growing at 22.3% CAGR. OpenAI, Anthropic, and Microsoft entering with dedicated legal verticals will accelerate commoditization of generic legal research — specialty WC/PI tools must differentiate on domain depth.
What it does: Superinsight.ai is a fully automated AI medical record review platform — no human reviewers involved — that produces structured medical chronologies, causation evidence analysis, ICD/CPT billing code extraction, ADL impairment documentation, and nexus letter support, with all findings precisely cited back to source pages. The platform is built for both plaintiff and defense work in the same system: for plaintiffs, it surfaces causation timelines and MMI documentation; for defense, it flags pre-existing conditions, treatment inconsistencies, and causation gaps.
WC/PI use case: Upload a 600-page workers' comp file and receive a structured causation analysis identifying which documented medical findings support AOE/COE, highlighting treatment appropriateness issues, and extracting impairment rating documentation — in hours rather than days, at a fraction of nurse reviewer cost. The system also supports IME physicians in organizing chronologies before issuing opinions.
Verdict: The no-human-reviewer architecture is both a selling point (privacy, speed, cost) and a caveat requiring careful attorney review of outputs — but for high-volume WC shops where causation is the primary battleground, the platform's focus on work-relatedness analysis makes it one of the most purpose-built tools in this category. → Workers' Comp Product Page
OpenAI's entry into legal product development — via Jason Boehmig's hire and the "Codex for Legal" brand — follows a clear agentic architecture playbook: foundation models connected to legal-specific retrieval systems, workflow agents for contract review and research, and enterprise integrations with existing practice management tools. Boehmig's stated mission is "building AGI for law," and his background at Ironclad (which built large-scale contract lifecycle management on GPT models) suggests the initial product focus will be document-centric: drafting, reviewing, and negotiating contracts and briefs at scale.
For WC/PI practitioners, the more immediate signal is what this does to the competitive dynamics of the legal AI vendor ecosystem. When OpenAI, Anthropic, and Microsoft each field dedicated legal verticals, they commoditize the generic layer — legal research, document drafting, general chat — which is exactly where most horizontal legal AI tools have been charging premium SaaS pricing. The durable value in this market shifts to domain depth: tools that understand IWCC arbitration procedure, Illinois PPD schedules, causation doctrine in occupational disease cases, or carrier-side RTW protocols. That's where Supio, DigitalOwl/ChartSwap, EvenUp, and InPractice have positioned themselves. The platform wars make domain-specific moats more valuable, not less.
Boehmig noted that "it's a mistake to believe any one player can do it alone, even a frontier lab" — an implicit acknowledgment that the legal AI market will remain partnership-driven, with OpenAI providing the model layer while specialist vendors own the workflow.
Ninth Circuit suspends two Orange County attorneys six months, fines each $2,500, and imposes two-year mandatory AI-disclosure requirement for filing briefs containing fabricated citations and misattributed quotations, compounded by repeated misrepresentations to the court. The opinion defines AI hallucination taxonomy (fabrications vs. inaccuracies) and issues bar-wide warning. Court cites 2024 study showing Westlaw/Lexis AI hallucination rates of 17%/33%. Full opinion available via Volokh Conspiracy.
Significance: Sets the firmest federal appellate AI-misconduct precedent to date; the cover-up aggravation factor will shape how courts evaluate subsequent violations.
Memphis law firm sanctioned by Judge Sheryl Lipman in the Western District of Tennessee for AI misuse in a malpractice suit against Baker Donelson, ordered to reimburse costs and report the matter to Tennessee disciplinary counsel. Details remain behind a Law360 paywall, but the action confirms the judicial sanctions trend is not geographically limited to the Ninth Circuit.
Significance: Second AI sanctions order in a single week from separate federal courts — WC/PI practitioners in all circuits should treat AI-output verification as non-negotiable court practice.
The Trump administration DOL's proposed rule to rescind the Biden 2024 independent contractor standard — which applied a multi-factor economic reality test — and revert to a two-factor economic dependence analysis is now in final-rule development. Unlike employees, independent contractors are not entitled to FLSA minimum wage, overtime, or workers' compensation coverage, making the classification standard directly material to WC coverage disputes in gig-economy injury cases.
Significance: A final rule expanding independent-contractor classification would shift WC exposure in transportation, warehousing, and food delivery cases; monitoring for finalization.
The Illinois Department of Human Rights announced June 2 that its public hearing on the proposed Subpart J regulations implementing HB 3773 / Public Act 103-0804 — the Illinois AI Employment Opportunity Act — has been "temporarily postponed" from its June 10 date to allow for "continued collaboration with other state agencies." The 45-day public comment window, triggered by the May 15 publication in the Illinois Register, remains open through June 29, 2026.
The proposed Subpart J rules require employers who use AI in covered employment decisions — recruitment, hiring, promotion, discharge, discipline, RTW determinations, and terms and conditions of employment — to provide written notice to employees that AI is being used. The law prohibits use of AI that discriminates based on protected characteristics even if the discrimination is unintentional, and bars use of zip codes as proxy for protected classes.
WC/PI exposure: The direct WC relevance is through RTW and accommodation decisions. If a carrier or TPA uses AI to evaluate an injured worker's fitness for return to work or to make job classification decisions that affect benefit eligibility, that system may trigger HB 3773 notice and non-discrimination obligations. Employers using AI-driven IME review tools, claims triage systems, or disability management platforms should assess whether those systems influence covered employment decisions. Violations expose employers to actual damages, civil penalties, attorneys' fees, and compliance reporting.
Monitoring: Next steps unclear — IDHR may revise the proposed rules before reopening the comment period. Stakeholders should monitor the IDHR Legislative Updates page and submit comments by June 29 if the window is not extended.
Azcon Metals v. IWCC, 2026 IL App (5th) 250301WC (Feb. 26, 2026): The Fifth District affirmed that a claimant who receives compensation under Section 8(e)'s scheduled-loss provision cannot also seek a wage differential award under Section 8(d)(1) — not "both at once." The decision restates the long-standing exclusivity rule between scheduled and unscheduled awards, but its recent citation frequency in briefs signals its continued use in permanent disability valuation disputes. Defense counsel handling high-wage claimants with scheduled injuries should review the Azcon rationale when opposing dual-award demands.
The AI medical record review market has matured from novelty to infrastructure in 2026, with platforms like InPractice, Superinsight.ai, DigitalOwl/ChartSwap Insights, Supio, and Tavrn each carving differentiated positions. But a candid practitioner guide from Quilia raises a distinction that should inform how WC/PI attorneys use these tools: AI excels at factual extraction (MRI findings, treatment dates, provider interactions) but cannot reliably perform legal analysis (whether a finding supports the demand, how a jury will receive it, whether it survives a Daubert challenge).
The guide flags a specific WC-relevant caveat: cases with older providers, handwritten notes, or non-standard formats see significantly reduced AI extraction accuracy — a real-world limitation for the large share of WC files that originate with occupational health clinics and community providers outside major hospital systems. The implication is that AI-generated chronologies in WC cases require more attorney-level verification than in standard auto tort files with modern hospital records.
On the IME side, Superinsight.ai has specifically built a product line for IME physicians and QMEs, helping evaluators organize chronologies and flag inconsistencies before issuing opinions. This raises a new evidentiary question beginning to surface in depositions: did the IME physician use AI assistance in reviewing records, and if so, what was the platform and what did it flag? Plaintiff counsel should begin asking this in IME deposition prep.
Thomson Reuters / Supio integration: The 2026 AI in Professional Services Report notes that the Thomson Reuters–Supio partnership is emerging as an enterprise-credibility signal for PI firms requiring institutional vendor validation alongside AI capability.
The full text of Wednesday's Ninth Circuit AI hallucination discipline order — available via the Volokh Conspiracy and Bloomberg Law Government — is the most important AI-in-practice document issued by any federal court in 2026. The opinion's taxonomy of fabrications versus inaccuracies, its detailed reconstruction of how the cover-up was detected, and its bar-wide warning language make it required reading for any attorney using generative AI in client work. Read it before your next brief. → Read at Volokh Conspiracy
The Ninth Circuit was explicit: lesser sanctions might have applied if the attorneys had disclosed AI use and apologized at the first opportunity. They didn't — and the escalation doctrine that court articulated is now available to every federal judge facing a similar pattern. In WC/PI practice, where AI is touching briefs, demand letters, interrogatory responses, and expert reports, the professional obligation to disclose and correct AI errors immediately is no longer an ethical preference. It's a case-management survival rule.
Jason Boehmig joining OpenAI to build "Codex for Legal" is not a threat to WC/PI-focused legal AI vendors. It's a signal that the generic layer of legal AI — research, drafting, document review — is being commoditized by platform players. That pressure actually helps the WC/PI-specific vendors who have built causation analysis, PPD calculation, RTW documentation, and IME review workflows. Domain depth becomes the moat. The attorneys who benefit most are those who already understand what the generic tools can't do.
The 9th Circuit and the W.D. Tennessee both acted on AI misconduct within hours of each other on June 3 — and these are not isolated outliers. The judicial enforcement curve is steepening. Firms that haven't formalized AI verification protocols, attorney-review requirements, and disclosure practices are accumulating professional responsibility exposure at an accelerating rate. If your firm doesn't have an AI-use policy with teeth by the time this edition publishes, write one today.
Illinois was the first state in the nation to enact a biometric privacy law — the Biometric Information Privacy Act (BIPA) in 2008 — which now serves as the legislative template for AI privacy and employment regulation like HB 3773. Illinois plaintiffs have recovered over $2 billion in BIPA settlements since 2019.
Fabrication (n., AI law context) — Per the 9th Circuit's LNU v. Blanche taxonomy: an AI-generated output citing cases or quotations that do not exist. Distinguished from "inaccuracy," where real authorities are cited but their holdings are misstated. The court identified both as sanctionable; inaccuracies may be more dangerous because they survive superficial citation checks.
"Lawyers using generative AI must thus be aware of the tendency of generative AI to make these mistakes and guard against them. When an attorney learns of any error in a filing — including generative AI hallucinations — he should immediately alert the court and opposing counsel of the error and disclose its source."
— Ninth Circuit, LNU v. Blanche, June 3, 2026