All the intelligence that's fit to argue.
The information asymmetry that has defined personal injury claims handling for decades — carriers armed with proprietary data, injured claimants armed with a lawyer or nothing — is collapsing faster than the profession anticipated. Mighty, a New York-based legal technology company, has completed the first motor vehicle injury claim settlements handled end-to-end by an AI agent, without a lawyer and without human involvement on Mighty's side. The platform — free to consumers — collects accident reports, police reports, and medical records, generates a real-time valuation, then deploys its AI agent to submit a demand package and negotiate directly with the carrier's AI system. If a settlement is reached, the consumer keeps 100% of the proceeds. If the case is too complex for pre-suit resolution, Mighty's agent escalates to a vetted attorney in its network — and that attorney's contingency fee doesn't kick in until they improve on the offer Mighty's AI already obtained. CEO Joshua Schwadron described the model plainly: "A lot of personal injury cases are straightforward and shouldn't have a lawyer taking 33–40% when there's no legal work to be done." This is the logical endpoint of a trend Legal Intel has tracked all year: EvenUp (now valued at over $2 billion) crowdsourcing settlement data from 2,000+ plaintiff firms; CLARA Analytics training carrier models on millions of closed claims; and now consumer-facing AI agents removing the attorney as intermediary entirely for a category of cases that, by volume, represents the bulk of the PI market.
Reporter's Take: The threat to PI practitioners isn't that Mighty is going to replace trial lawyers. It's that Mighty correctly identifies the category of cases — the 90% of PI matters that settle pre-suit without meaningful legal work — and strips the contingency fee from them. The 10% of cases requiring genuine legal strategy become the entire market for PI attorneys going forward. Firms that can't articulate their differentiated value in that 10% are in structural trouble.
The Texas Supreme Court agreed June 5 to hear a dispute over whether a nonstenographic deposition transcript generated using AI-driven voice recognition technology can be used in litigation, after a lower court struck the transcript and barred future depositions using the same method. The case raises the question directly: does AI transcription satisfy Texas's long-standing authorization of non-stenographic deposition recording, or does it introduce reliability problems that require per se exclusion? Why it matters: For WC/PI practitioners, AI deposition transcription is already in wide use via platforms like Skribe, Prevail, and Lexitas's Deposition Insights+. A Texas Supreme Court ruling setting exclusion standards will set a benchmark for evidence foundation challenges in other jurisdictions and directly affect cost-of-deposition math in high-volume WC litigation.
Source: Law360 (June 5, 2026)Federal courts issued four significant and diverging rulings in Q1 2026 on whether AI-assisted litigation preparation is protected by attorney-client privilege or the work-product doctrine. The central split: Warner v. Gilbarco (E.D. Mich., Feb. 10, 2026) protected a pro se plaintiff's ChatGPT-assisted litigation materials as work product; United States v. Heppner (S.D.N.Y., Feb. 17, 2026) held that a criminal defendant's solo Claude AI sessions were not privileged because the tool was a third party and the communications weren't at counsel's direction. In June, the Texas Business Court added a nuanced middle position: non-lawyer principal's ChatGPT conversations protected as work product, but disclosure of which discovery materials were fed to the tool required. Why it matters: How attorneys and clients use AI — specifically whether AI is deployed at counsel's direction as part of the legal strategy — is now the decisive privilege variable. WC/PI firms need an AI-use policy before a discovery motion forces one.
Source: National Law Review / Akin Gump / White & CaseColorado's original AI employment discrimination law is effectively a dead letter: a federal court injunction entered April 27, 2026 halted enforcement days before the June 30 effective date, and Governor Polis has signed SB 26-189 — a significantly narrower replacement that takes effect January 1, 2027. The new law focuses on transparency and notice requirements for automated employment decision tools, removing the broad risk management and impact assessment requirements of the original. Why it matters: Carriers and employers using AI for WC return-to-work and fitness decisions in Colorado face a one-year window of reduced compliance burden before the new framework activates. Use it to build audit documentation infrastructure voluntarily — discovery in future contested claims will ask whether you had it.
Source: Littler / National Law ReviewPrivate equity firms are openly questioning the durability of BigLaw's hourly billing model as AI adoption accelerates, according to reporting in Law360. PE clients are benchmarking legal work against AI-capable alternatives and pushing for hybrid value-based pricing. Why it matters: For WC/PI defense firms on carrier panel, the same logic applies — carriers holding AI-generated cost benchmarks for standard defense tasks will use them in panel contract negotiations. Defense firms that can't demonstrate AI-augmented efficiency will face fee compression at renewal.
Source: Law360Governor Pritzker has not yet signed SB 315 as of June 8, though he has publicly committed to doing so. The legislature has 30 days to transmit the bill formally; once signed, the Artificial Intelligence Safety Measures Act takes effect January 1, 2027 — not 2028 as previously reported in some outlets. The effective date correction matters: frontier AI developers building WC/PI tools on large models have roughly six months to prepare their audit-readiness infrastructure. Why it matters: Illinois practitioners advising carrier clients on AI vendor contracts should begin requiring audit-readiness representations in agreements with frontier-model-based tool providers now, before the January 2027 compliance window forces renegotiation.
Source: Airia / IAPP (June 2026)The Mighty model — consumer AI negotiating against carrier AI — raises a foundational question that neither the insurance industry nor legal technology has yet answered: what happens when two proprietary AI models reach opposing valuations and neither will move? Dan Ivtsan, writing in Insurance Thought Leadership, framed it directly: does a shared analytical baseline accelerate resolution because both sides converge on fair value, or does it entrench positions because each side treats its own model as authoritative truth?
The data divergence problem is structural. CLARA Analytics trains on carrier-side closed claims data — a dataset optimized around how carriers historically resolved claims, not how courts awarded them. EvenUp trains on plaintiff attorney settlement data — a dataset filtered through the negotiating behavior of plaintiff firms, which tends to anchor higher. When these two data architectures produce conflicting valuations and the negotiating agents both have authority to hold firm, the "straightforward" pre-suit case can deadlock faster than human negotiators would.
For WC/PI builders: The next infrastructure gap in the market is a trusted neutral AI valuation layer — a third-party model trained on actual verdict and resolution data, not party-specific negotiating data, that both sides can reference as a settlement anchor. This is the ADR infrastructure play of 2026–2027. The firm or startup that builds a defensible neutral AI verdict/settlement database for WC/PI claims will have leverage across both sides of the bar.
The Texas Supreme Court's acceptance of the AI deposition transcript case forces a direct confrontation with the evidentiary reliability of AI voice recognition in legal proceedings. The stakes are significant for WC/PI practice: AI transcription platforms like Skribe, Prevail, and Lexitas Deposition Insights+ have gained wide adoption because they dramatically reduce deposition costs — in high-volume WC litigation where depositions of treating physicians, vocational experts, and IME doctors generate thousands of pages of transcript annually, cost compression is real. But if the Texas Supreme Court holds that AI transcripts require a foundation of human verification or per se fail the accuracy standard, the cost savings evaporate and the deposition service market resets. Watch this closely. A ruling restricting AI transcription will be cited in federal courts nationwide before the ink is dry.
The structural problem with AI-to-AI PI settlement negotiation is that both models are trained on their own side's historical data. Carrier AI (CLARA, Gradient AI) is trained on closed-claims data that reflects how carriers have historically resolved claims — which includes the effect of prior information asymmetry and claimant underrepresentation. Plaintiff AI (EvenUp, Mighty) is trained on attorney-mediated settlement data anchored by firm-side demand patterns. Neither reflects actual jury verdict distributions for comparable injuries. The result: automated agent negotiations may converge on historically distorted "market" values rather than legally correct compensation. For plaintiff WC/PI attorneys, the response is to hold jury verdict data as an independent reference point — it's the one dataset neither carrier AI nor pre-suit settlement AI is trained on.
As carriers deploy AI-driven social media surveillance and activity monitoring for WC fraud detection, the NAIC Model Bulletin's bias testing requirements create an emerging obligation: carriers must test whether their surveillance AI flags claimants at disparate rates based on protected characteristics. An AI fraud detection model trained on historical claims data will inherit the racial, gender, and age biases embedded in historical adjuster decisions. Plaintiff attorneys should be asking in discovery not just "did you surveil my client with AI" but "what bias testing has been performed on your surveillance model and what were the results."
The most sharply-framed analysis of the AI data-asymmetry collapse in PI/WC claims published this cycle. Ivtsan maps the competitive dynamics between CLARA Analytics (carrier-side), EvenUp (plaintiff-side), and Mighty (consumer-side), and asks the question the industry isn't answering: when both models disagree, what resolves the dispute? The piece is essential reading for any WC/PI practitioner trying to understand how AI is changing the negotiating landscape — and for any legal tech founder thinking about where the neutral infrastructure gap is.
Four diverging federal court rulings in four months. The courts are working out the AI work-product doctrine in real time, and the variable that determines the outcome in every case is the same: was the AI used at counsel's direction, integrated into legal strategy, or deployed independently by the client? The practical implication is that attorney-directed AI use needs to be documented — not just performed. An email from counsel directing a paralegal to run medical records through an AI chronology tool, retained in the file, is a better evidentiary record than the same task performed without documentation. This is litigation hygiene, not paranoia.
When a free AI agent can value and settle a straightforward motor vehicle injury claim with zero legal fees to the consumer, the contingency fee structure is revealed for what it always was: a pricing model for access, not a pricing model for legal complexity. The cases Mighty targets — the simple rear-enders, the soft-tissue claims with clear liability — were always administrative work priced as if they were legal work. The profession has known this for decades; Mighty is just the first product that makes the arbitrage explicit to consumers. The correct response is differentiation, not denial.
With SB 315 taking effect January 1, 2027, Illinois has a six-month window to define implementation — audit standards, incident reporting protocols, attorney general enforcement priorities — before any other state has to follow. The IAPP's assessment that Illinois, Connecticut, and New York are moving in concert suggests a tri-state governance cluster is coalescing around frontier AI oversight. If those three states align their audit standards, the "Sacramento Effect" applies: major AI developers will audit to the highest common standard, and that standard becomes national by default. Illinois practitioners should be engaged in that standard-setting now.