Legal Malpractice Carriers Face First Wave of AI Claims as LPL Coverage Forms Outpace the Cyber Cycle

From monitoring coverage form language across 13 LPL carriers over the past two years, the trajectory is unmistakable: legal professional liability has moved from total silence on AI to active form modification faster than any prior coverage evolution in specialty lines. EPIC Insurance Brokers’ 16th Annual Lawyers’ Professional Liability Claims Survey, released in May 2026, delivers the first credible loss emergence data for AI-related malpractice. Seven of 13 surveyed carriers report an increase in AI-related claims over the past year. This is no longer a hypothetical risk category. LPL actuaries now confront a pricing problem with no development history, coverage forms in mid-revision, and a claims pipeline being fed by over 1,200 documented cases of attorneys submitting AI-fabricated citations to courts worldwide.

The EPIC Survey: First Credible Loss Emergence Data

EPIC’s annual survey covers 13 LPL carriers that collectively insure more than 80% of Am Law 200 firms, making it the most comprehensive barometer available for lawyers’ professional liability trends. The 2026 results mark several firsts that LPL pricing actuaries should be tracking closely.

Seven of 13 carriers (54%) reported AI-related claims increases over the past year. This is the first time a credible survey has produced loss emergence data specific to AI in any professional liability line. The finding transforms the conversation from “will AI create malpractice claims?” to “how fast is the frequency developing?”

Eight of 13 carriers (62%) reported higher overall claim frequency, the first rise in five years across the LPL market. Whether the AI claims component is driving the broader frequency increase or merely coinciding with it remains an open question, but the correlation is difficult to dismiss.

Eleven of 13 carriers (85%) reported materially higher defense spending year-over-year. AI-related claims are disproportionately expensive to defend because they often involve novel questions of professional competence, technology-specific discovery, and multiple proceedings (the malpractice suit, the underlying case, and potential bar disciplinary action) running in parallel.

Nine-figure claims ($100 million and above) are “no longer viewed as rare outliers,” according to the survey. EPIC’s historical data across $5.2 billion in analyzed gross written premiums from 2006 to 2018 shows $7.8 billion in severe losses at the $5 million-plus threshold, with accident year 2018 alone producing one claim over $100 million and multiple claims exceeding $10 million.

Eileen Garczynski, Principal of EPIC’s Law Firm Group, summarized the shift: “AI-related malpractice exposure has moved from theoretical to real. The duty of competence cannot be delegated to technology.”

The broader LPL market context from AM Best reinforces why this matters for pricing. Direct premiums written for the AM Best LPL composite of 16 specialty insurers reached $728 million in 2025, up from $709 million in 2024. Cumulative growth exceeded 18% since 2020, a meaningful acceleration after the near-flat growth of less than 1% between 2015 and 2019. The operating ratio for the composite was 58.7 in 2024, significantly outperforming the 84.7 commercial casualty composite. That margin cushion partly explains why no major LPL writer has yet filed an explicit AI exclusion: the line has been profitable enough to absorb emerging AI losses without immediate rate action. But the EPIC frequency data suggests that absorption capacity may be tested sooner than current rate plans contemplate.

Court Sanctions Build the Claims Pipeline

The claims pipeline feeding LPL carriers draws from a growing database of court sanctions for AI-fabricated legal work product. Damien Charlotin’s Hallucination Database at the HEC Paris Smart Law Hub documents 1,227 cases globally as of early 2026, with 811 in the United States. The growth rate is accelerating: from approximately 200 cases one year ago to 719 in January 2026 to over 1,227 now, averaging five to six new documented cases per day.

The composition of the database reveals a concentrated risk profile. Of the 1,227 documented incidents, 1,022 involved fabricated case citations and 323 involved false quotes from real cases. Solo practitioners account for 50.4% of all cases, with small firms under 25 attorneys representing 89.9% of documented incidents. From a claims severity perspective, this concentration in small firms means the risk falls disproportionately on carriers writing smaller accounts, where per-policy defense costs represent a larger share of premium.

Case	Court	Date	Details	Sanctions
Mata v. Avianca, Inc.	S.D.N.Y.	June 2023	Six fabricated cases from ChatGPT	$5,000 fine
Park v. Kim	Second Circuit	2024	AI-generated fake citations	Referral for discipline
Kruse v. Karlan	Missouri Court of Appeals	2024	Multiple AI-fabricated citations	Appeal dismissed
Whiting v. City of Athens	Sixth Circuit	March 2026	24+ fake citations, defied show cause order	$15,000 per attorney + full fee reimbursement + double costs
Mississippi dual-side case	Federal court, Mississippi	November 2025	Both sides independently submitted AI-fabricated briefs	$3,500 + 2-year bar (defense); $2,500 + CLE (plaintiff)

The largest single penalty on record stands at $109,700. But direct sanctions understate the total exposure because each incident can generate multiple follow-on claims: a malpractice suit from the affected client, disciplinary proceedings from the state bar, and reputational harm that compounds across a firm’s entire book of business. From tracking these cases over two years, patterns in how we’ve seen these claims develop suggest that the follow-on malpractice claims typically emerge six to eighteen months after the initial court sanction, creating a lag that complicates reserving.

Coverage Form Evolution: From Silence to Active Modification

Pre-2023 LPL policies contain no AI-specific language. Coverage for AI-related claims under these forms depends entirely on how carriers interpret existing provisions: the professional services definition, intentional acts exclusions, breach of confidentiality clauses, and unauthorized practice restrictions. The market has moved through four distinct phases since 2023, a pace of evolution that parallels and now exceeds the CGL AI exclusion development cycle.

Phase 1: Silent cover (pre-2024). The vast majority of LPL policies in force today were written without any AI-specific language. These forms neither affirm nor exclude AI-related claims, leaving coverage determination entirely to the carrier’s interpretation at claims time. For pricing actuaries, this means the existing book contains unquantified AI exposure that is not reflected in historical premium levels or rating algorithms.

Phase 2: Underwriting questionnaires (2024-2025). CNA, the largest U.S. legal malpractice carrier, added supplemental AI questionnaires to its renewal process starting in 2025. CNA distributed a March 2025 ethics bulletin and identified AI as a novel claims exposure in its fiscal year 2025 10-K filing. AmTrust has the only confirmed verbatim AI question on a U.S. LPL application form (LPLPRO-APP-01 0523): “Allow the use of Artificial Intelligence software to draft documents. If checked, please attach description.” Per Aon SVP Stan Sterna (April 2026), standard LPL underwriter questions now include: “Do you use AI? Do you police it? Do you have protocols in place?”

Phase 3: Exclusionary or governance-contingent (2025-2026). Hamilton Select has filed the broadest known AI exclusion in any professional liability form: “any claim, wrongful act, damages, or defense costs based upon, arising out of, or in any way involving any actual or alleged use of generative artificial intelligence by the insured.” The endorsement names ChatGPT, Bard, Midjourney, and DALL-E by product. W.R. Berkley’s Form PC 51380 applies an “absolute” AI exclusion to management liability lines (D&O, EPL, fiduciary), though not LPL-specific forms. On the general liability side, ISO/Verisk’s CGL endorsements (CG 40 47, CG 40 48, CG 35 08) took effect January 1, 2026, and by April 2026, W.R. Berkley, Chubb, Travelers, Berkshire Hathaway, and Cincinnati Financial had filed to adopt these or proprietary AI exclusion language with over 80% of state filings approved.

Phase 4: Affirmative but conditioned cover (2025-2026). A parallel market for explicit AI liability coverage is emerging. Armilla AI, backed by Lloyd’s syndicate capacity through Chaucer and Axis Capital, offers $25 million limits and expanded its program in February 2026, explicitly referencing lawyer hallucination cases as a target risk. Counterpart launched affirmative AI coverage on miscellaneous professional liability and Tech E&O forms in November 2025. HSB (Munich Re) introduced AI liability coverage for SMEs in March 2026. Munich Re’s aiSure program provides performance-warranty coverage on a case-by-case underwriting basis. Beazley has publicly stated it has no plans to add AI exclusions to its professional liability forms.

A critical distinction for LPL actuaries: as of May 2026, no major U.S. LPL writer (CNA, Travelers, Chubb, Hartford, Markel, ALPS, or state mutuals) has filed an explicit AI exclusion on its named lawyers’ professional liability form. AI exclusions in the professional lines space are concentrated on management liability and general liability. The LPL line remains largely in Phase 2, with carriers gathering data through questionnaires rather than restricting coverage through endorsements. Nine carriers surveyed by Legal AI Governance (Aon Attorneys’ Advantage, Berkley Select, LMIC California, MLM, OAMIC, Old Republic Pro, Bar Plan Mutual, Travelers, and WILMIC) have no public AI position at all.

What Carriers Ask at Renewal

Over 60% of LPL carriers now include AI-related questions in their intake and renewal applications. WTW identified the period from January 2025 to January 2026 as a “structural break” in the professional liability market regarding AI, with silence about AI in renewal documentation largely replaced by underwriter questions, governance-document requests, and manuscript exclusions.

CNA’s supplemental AI questionnaire covers five core areas: which AI tools the firm uses, whether a written AI governance policy exists, whether attorneys receive AI-specific training, how AI-generated work product is reviewed and verified, and whether the firm has an AI incident response plan. Firms that cannot demonstrate governance may face higher premiums or restrictive endorsements at renewal.

Carriers are increasingly expecting a renewal-ready documentation package mapped to ABA Formal Opinion 512 (July 2024), the first comprehensive national ethics framework for attorney AI use. The nine items carriers are requesting include: a written firm AI policy dated within 12 months, a vendor due-diligence file, tool-specific informed-consent language, training and CLE records, a pre-filing verification log, a usage log or register, an incident response procedure, quarterly managing-partner attestation, and a renewal-ready audit report.

The regulatory backdrop compounds the underwriting signal. ABA Opinion 512 establishes that lawyers can use AI but their ethical obligations under Model Rules 1.1 (Competence), 1.6 (Confidentiality), 1.4 (Communication), and 5.1/5.3 (Supervision) do not change. Thirty-five state bars have now issued AI-specific guidance, and 25-plus federal district courts have adopted standing orders requiring attorneys to certify whether AI was used in filings. The Embroker Legal Industry Risk Index reports that AI adoption among law firms grew from 58% to 80% year-over-year, while 44% of firms still lack formal AI governance policies according to the Clio Legal Trends Report. That gap between adoption and governance is the risk that carriers are now pricing through questionnaires, because they lack the loss data to price it through actuarially derived rate factors.

The Actuarial Pricing Problem: Building Rates Without Loss Triangles

Insurance pricing rests on three foundations: historical loss data, predictable causation, and defined coverage scope. AI-related legal malpractice challenges all three simultaneously, creating a pricing problem unlike anything LPL actuaries have faced since the emergence of cyber exposure in the early 2000s.

No historical loss development data. AI-related legal malpractice claims are too new to have credible loss triangles. The earliest documented sanctions case (Mata v. Avianca) dates to June 2023. Even if every sanctioned case generated a malpractice claim immediately, which they have not, actuaries would have at most three years of accident-year data with no tail development factors. Chain-ladder, Bornhuetter-Ferguson, and other standard reserving methods require minimum development periods that do not yet exist for this exposure class.

Unpredictable causation. The underlying technology evolves faster than legal and regulatory systems can respond. A model that produces fabricated citations in 2024 may not produce them in 2026 following architectural improvements, or it may produce a different category of error entirely. The causal mechanism shifts with each model generation, making frequency assumptions derived from current data unreliable for even near-term projections.

Undefined coverage scope. The spectrum of coverage responses across the 13 surveyed carriers means that two similarly situated law firms with different carriers can face entirely different coverage outcomes for identical AI-related claims. One firm’s carrier may affirm coverage under a governance-contingent endorsement while another’s may deny it under a silent policy form. This heterogeneity makes industry-level loss data, even once it develops, difficult to apply across books of business.

Current market pricing provides rough benchmarks. AI-specific endorsements cost 5% to 15% above base premium, adding $2,500 to $7,500 for a mid-size firm paying $50,000 per year. A single AI-related malpractice event at a mid-size firm can generate $500,000 to $2 million in defense costs and damages. Generative AI lawsuits rose 978% from 2021 to 2025, with over 700 cumulative cases filed. Gartner projects over 2,000 “death by AI” legal claims by the end of 2026.

From tracking rate filings across specialty lines, the approach most LPL actuaries are taking mirrors what NAIC regulators are observing in other AI-affected lines: borrowing frequency and severity assumptions from analogous emerging risk categories. Cyber liability in the 2005-2010 period is the closest precedent, where carriers used executive judgment overlays on thin data to set initial rates that were later refined as credible development emerged. The risk of this approach is well understood: if initial assumptions understate the true frequency or severity, the book builds in underpricing that compounds across renewal cycles before credible data forces corrections.

Faster Than Cyber: Comparing the Coverage Evolution Timelines

Timothy Zeilman, Global Head of Product Ownership at HSB (Munich Re), frames the comparison directly: “We’re seeing the same pattern we saw with cyber 15 or 20 years ago.” The parallel is instructive, but the pace is different.

Cyber insurance followed a roughly 20-year arc from emergence to maturity. Silent cover embedded in existing policies defined the 2000 to 2005 period. Standalone cyber products and explicit exclusions on non-cyber policies emerged between 2010 and 2015. By 2020, cyber was a mature standalone line with standardized forms, actuarially credible loss data, and dedicated pricing models.

AI liability coverage is compressing that timeline dramatically. The silent cover period (pre-2024) lasted roughly two years from the release of ChatGPT in November 2022, compared to five-plus years for cyber. Questionnaires and exclusions arrived within 18 months (2024-2026), compared to approximately a decade for cyber. Affirmative standalone products (Armilla AI, Counterpart, HSB) appeared within three years of the first publicized incident, compared to 10-plus years for cyber standalone policies. If this trajectory holds, the LPL market could reach coverage standardization for AI risks by 2028 or 2029, a fraction of the cyber timeline.

Phase	Cyber Coverage Timeline	AI Liability Coverage Timeline
Silent cover	~2000-2005 (5+ years)	~2022-2024 (2 years)
Questionnaires and exclusions	~2005-2015 (10 years)	~2024-2026 (2 years)
Standalone affirmative products	~2010-2015 (10+ years from emergence)	~2025-2026 (3 years from emergence)
Mature standardized market	~2020 (20 years from emergence)	Projected 2028-2029 (6-7 years)

Two structural differences explain the acceleration. First, the insurance industry already has the institutional vocabulary and market infrastructure for coverage innovation from the cyber experience. The playbook of questionnaire, exclusion, affirmative product, and standardization exists and can be executed faster the second time. Second, AI risks are embedded across multiple lines simultaneously (professional liability, D&O, general liability, cyber), not concentrated in a single line as cyber initially was. This cross-line exposure creates competitive pressure to act because carriers that remain silent on AI in their LPL forms face adverse selection from risks that have been excluded elsewhere. HSB data shows that 74% of small businesses currently use AI tools and 91% expect to adopt AI in the near future, meaning the exposed population is far larger and more diffuse than the early cyber-exposed population was.

Why This Matters for Actuarial Practice

The convergence of loss emergence data, coverage form evolution, and regulatory pressure produces five specific implications for actuarial work across professional liability lines.

Loss development factor selection for silent AI claims already in the pipeline. Every LPL policy written before 2024 contains unpriced AI exposure. As claims develop against these policies, actuaries selecting LDFs for recent accident years must decide whether AI claims represent a structural shift in development patterns or a temporary frequency spike that will stabilize once coverage forms catch up. The EPIC data showing 54% of carriers reporting increases, combined with the 5-6 cases per day feeding the hallucination database, argues for the structural interpretation. LDF selections that rely on pre-2023 development patterns will likely understate ultimate losses for 2024 and 2025 accident years.

Rate adequacy for AI-conditioned coverage. Carriers moving to governance-contingent coverage (Phase 4) need rate indications that reflect the risk differential between firms with robust AI governance and those without. The 5% to 15% surcharge currently applied to AI endorsements is not actuarially derived; it reflects underwriting judgment in the absence of loss data. As credible data develops, these initial assumptions will need to be tested against actual experience, and the risk of initial underpricing in a line with no tail data is substantial.

Reserving for claims already filed under pre-AI policies. The EPIC survey confirms that AI-related claims are already being reported under silent policies. Case reserves for these claims cannot rely on historical benchmarks for legal malpractice defense costs because AI cases involve novel discovery, technology-specific expert witnesses, and multiple parallel proceedings. Defense cost severities for AI-related claims are likely to exceed the per-claim averages embedded in current reserve positions. The pattern already visible in medical malpractice, where verdict severity has doubled in two years, offers a cautionary precedent for professional liability lines absorbing new risk categories.

Cross-line implications as the LPL pattern spreads. Legal malpractice is the first professional liability line to produce credible AI loss data because lawyers were among the earliest professional adopters of generative AI and because courts provide a public record of failures. But the pattern will extend to other professional lines. Medical malpractice carriers face diagnostic AI errors. Accounting firms face AI-assisted audit failures. Engineering and architecture firms face AI-generated design flaws. Each professional liability line will eventually encounter the same pricing problem that LPL actuaries face today, and the LPL experience is building the precedent for how coverage forms and rate structures respond. The litigation funding dynamics that are already amplifying severity across tort lines will accelerate this cross-line spread, as funders identify AI-related professional liability claims as a new category with asymmetric information advantages for plaintiffs.

ABA Opinion 512 creates an underwriting bright line. The ABA’s formal opinion establishes a duty-of-competence floor that carriers can use to differentiate risk. A law firm that can document compliance with ABA Opinion 512 requirements (written AI policy, verification protocols, training records, supervision procedures) is a fundamentally different risk than one that cannot. This bright line enables actuarial segmentation once sufficient data accumulates, but it also creates a potential E&O trap: carriers that continue to write silent coverage for firms without documented governance may face challenges defending claim denials when the industry standard has shifted to requiring governance documentation at renewal.