W.R. Berkley has attached an absolute AI exclusion to its directors and officers, errors and omissions, and fiduciary liability forms, a scope none of the ISO general liability endorsements reach, and S&P Global Market Intelligence counted 41 P&C groups that have filed at least one subsidiary to exclude generative AI, with 20 more filing to delay (The Insurer, July 23, 2026). The question the move raises is not whether carriers will keep excluding AI. It is whether a management liability book can price a tail it has just declared it will not model.
A Filing Wave That Started in GL and Just Widened
The AI exclusion wave traces to a single filing. Verisk's Core Lines Services unit made a multistate submission in July 2025 that included an optional generative AI exclusion for commercial general liability, and the resulting endorsements, CG 40 47, CG 40 48, and CG 35 08, began attaching to CGL renewals effective January 1, 2026. CG 40 47 strips both bodily injury and personal and advertising injury coverage for any loss arising out of generative AI. CG 40 48 removes only the personal and advertising injury piece. CG 35 08 reaches products and completed operations. All three share the same broadly construed trigger, "arising out of," which requires only a causal link to generative AI rather than direct causation, and this publication's earlier coverage of the 80% state approval rate on those CGL filings traced how quickly that language cleared state review.
The Insurer's July 2026 analysis, built by running nearly 10,000 filings through S&P Capital IQ's product filings search, is the first count of how far the wave has spread across the market rather than just naming individual carriers (The Insurer, July 23, 2026). Forty-one groups with a subsidiary that filed an exclusion, against 20 more that filed specifically to delay, is a market still mid-decision: a plurality of large carrier groups has moved, but a meaningful cohort is deliberately holding off, which is itself informative. A carrier that files to delay is choosing to keep AI-related claims inside its existing book for another renewal cycle rather than carve them out immediately, a wager that either the loss experience will clarify before the exposure grows much larger, or that a later, better-negotiated exclusion is worth the interim risk.
Layered on top of the filing count is the litigation trend the exclusions are responding to. Generative AI-related lawsuits in the United States grew 978% from 2021 to 2025, with the year-over-year filing increase accelerating to 137% between 2024 and 2025 (Carrier Management, July 23, 2026). Joe Lam, Verisk's vice president of liability, framed the underwriting logic behind the standardized forms directly: "Without exclusions to allow underwriters a level of stability or to accept a risk, you run into a situation where they might just walk away from the risk" (Carrier Management, July 2026). That is the GL story. The management liability story is different, because Berkley's form does not stop where the ISO forms stop.
Berkley's PC 51380 Reaches Where the CGL Forms Never Did
W.R. Berkley's "Artificial Intelligence Absolute Exclusion," filed as Form PC 51380, bars coverage for any claim "based upon, arising out of, or attributable to" the actual or alleged use, deployment, or development of artificial intelligence, by any person or entity, and it applies across the carrier's D&O, E&O, and fiduciary liability products (National Law Review, May 2026). Bloomberg Law's July 2026 reporting called it insurance's "first-of-its-kind" absolute exclusion reaching into board-level coverage (Bloomberg Law, July 2026). The scope is the point. CG 40 47 and its siblings exclude bodily injury, property damage, and personal and advertising injury arising from generative AI in a commercial general liability context: a defective product, a defamatory chatbot, a factory injury. PC 51380 excludes a different category of harm entirely: shareholder derivative suits over inadequate AI governance, securities claims alleging misrepresented AI capability, breach-of-fiduciary-duty claims tied to AI-driven plan administration decisions, and, notably, a company's own failure to detect or disclose third-party AI use in its filings and communications. A GL exclusion protects a carrier from a physical or reputational event caused by AI. A management liability exclusion protects a carrier from a board's own decisions about AI, which is a fundamentally different, and far more litigated, category of claim.
| Form | Lines covered | What triggers the exclusion |
|---|---|---|
| CG 40 47 / CG 40 48 / CG 35 08 (Verisk/ISO) | CGL, products/completed operations | Bodily injury, property damage, or advertising injury arising out of generative AI use |
| PC 51380 (W.R. Berkley) | D&O, E&O, fiduciary liability | Any claim arising out of AI use, deployment, development, governance failure, or non-disclosure, regardless of whether AI caused the underlying harm |
That last clause, inadequate AI governance as an independent trigger, is the one actuaries pricing or reserving a management liability book should read most carefully. It means the exclusion does not require proof that an AI system malfunctioned or produced a bad output. A board that approved an AI strategy without documented oversight, and is later sued for it, can lose D&O coverage under PC 51380 even if no AI model ever generated a defective decision. Reed Smith partner Courtney Horrigan, representing policyholders, put the practical alarm plainly: "Given that everything including our phones with Google now uses AI, I don't know how anything would be left if there is a broad AI exclusion" (Bloomberg Law, July 2026). Lathrop GPM's Alana McMullin made the same point from the coverage-gap side: "There's been a major shift in the insurance industry's treatment of AI-related risks and insurers are moving very quickly to limit this exposure" (Carrier Management, July 2026).
Why a D&O Loss Triangle Does Not Look Like a GL Triangle
General liability claims from AI use tend to be idiosyncratic: one chatbot, one defamatory output, one injured party, reported and often resolved within a few years of the policy period. That is the loss pattern the CGL exclusions were built to carve out, and it is why the reserving conversation around those forms has mostly been about the size of the coverage gap, not the shape of the loss development. Management liability claims behave nothing like that. Securities class actions and fiduciary breach suits are long-tail, frequently taking three to seven years from the underlying event to final resolution once discovery, class certification, and appeals run their course, and they are correlated rather than idiosyncratic: a single disclosure about a company's AI capabilities can generate one claim naming dozens of directors and officers, with damages scaling to the entire shareholder class rather than a single injured party.
Cornerstone Research's midyear count makes the correlation and severity concrete. Securities class action filings rose 30% to 121 in the first half of 2026 compared with the prior six months, and 15 of those, 13% of core filings, were AI-related (Cornerstone Research, July 29, 2026). That 13% of filings carried $385 billion of the $529 billion disclosure dollar loss index for the half, 73% of the total. Stanford Law professor and former SEC commissioner Joseph Grundfest summarized the asymmetry: "AI-related cases represented a modest share of total filings but an outsized share of alleged investor losses in the first half of 2026" (Cornerstone Research, July 29, 2026). A frequency-modest, severity-dominant claim class is exactly the profile that breaks a loss-cost model built on GL-style idiosyncratic events; it is also exactly the profile a D&O actuary already knows how to reserve for from prior securities litigation cycles, tech-bubble and financial-crisis filings among them, which is the argument for pricing this exposure rather than excluding it outright.
The Exclude-or-Price Decision and What It Assumes About Severity
Not every carrier is choosing Berkley's path. McMullin's second observation, less quoted than her comment on the pace of change, is the one that frames the actuarial choice directly: "We may also see insurers who take advantage of the uncertainty in the market...choose to underwrite and price the risks probably for an additional premium rather than exclude them" (Claims Journal, July 2026). An absolute exclusion is, in effect, a statement that the carrier cannot bound the tail well enough to hold capital against it at any price the market would pay. Pricing the exposure for additional premium is the opposite statement: that the tail is wide but bounded, and that a loaded rate, however imprecise in year one, is preferable to ceding an entire growing line of business to standalone AI liability writers.
The broader D&O market context makes that second path look less far-fetched than it might in isolation. Public-company D&O rates rose a modest 2% to 5% in 2025 even as the overall market stayed soft, while private D&O held flat to down 5% on more than half of renewals (Founder Shield, January 2026). That small pocket of hardening in an otherwise soft market lines up with underwriters already treating AI-linked governance and disclosure risk as a rateable factor rather than an unrateable one, exactly the assumption an exclusion forecloses. A carrier writing PC 51380 into a renewal is not just avoiding a bad loss year; it is exiting the segment of the book where rate is currently moving in the carrier's favor, in exchange for certainty that no AI-labeled claim reaches the policy at all.
IBNR Exposure When the Exclusion Gets Litigated
The reserving problem does not end once a carrier decides to exclude. Broad, novel exclusionary language invites coverage litigation over what "arising out of" AI actually means for a specific claim, and every one of those disputes is a source of uncertainty for current accident-year loss picks on business written before the exclusion attached. A director sued over a decision informed by, but not solely caused by, an AI-assisted analysis presents exactly the ambiguous fact pattern PC 51380's broad triggering language was written to sweep in, and exactly the fact pattern a court applying the "arising out of" standard broadly, as it has for decades in general liability coverage disputes, might also sweep in on the insurer's side. Where a court instead narrows the phrase, coverage attaches on a claim the carrier's pricing never contemplated as an AI risk in the first place, since it was written before AI became a named peril.
For a reserving actuary, that ambiguity sits directly in IBNR. Policies written in 2024 and 2025, before any AI exclusion attached, remain exposed to AI-related D&O and fiduciary claims that have not yet been reported, given the multi-year lag between an alleged governance failure and a filed securities suit. Those claims will emerge against expired policies carrying no AI exclusion at all, meaning the "silent AI" exposure this publication has tracked in general liability books has a management liability analogue that is larger in dollar terms and slower to surface, given the years-long gap between an AI-related board decision and the securities suit or derivative claim it eventually produces. A reserving actuary building IBNR for 2024 and 2025 D&O and fiduciary accident years cannot rely on the exclusion to bound that exposure, because the exclusion was not in force when those policies were written; the tail sits entirely inside existing loss development factors that were never calibrated to an AI-correlated claim class, and the Cornerstone data on how disproportionately severe that class already is should be treated as a live input into current-year selections, not a future-state hypothetical.
Further Reading
- Verisk CG 40 47 Creates an AI Liability Pricing Gap: the ISO endorsement mechanics and GL loss-load adjustment methodology this article's D&O comparison builds on.
- CGL AI Exclusions Win 80% State Approval as Carriers Shed Generative AI Risk: the state regulatory reception that let the GL exclusion wave move this fast.
- GenAI Lawsuits Surge 978% With No Actuarial Pricing Baseline: the litigation growth data behind this article's frequency discussion, mapped into pricing terms.
- The Deploy-and-Exclude Paradox: how carriers writing AI exclusions are simultaneously the heaviest internal AI adopters.
- AI Risk Selection in Commercial Lines Submissions: how AI adoption signals are already entering underwriting decisions on the accounts these exclusions attach to.
Sources
- The Insurer: More than 60 P&C insurance groups file to adopt AI exclusions (July 23, 2026)
- Carrier Management: Insurer Interest in AI Exclusions Growing as Risk Becomes Omnipresent (July 23, 2026)
- Claims Journal: Insurer Interest in AI Exclusions Growing as Risk Becomes Omnipresent (July 20, 2026)
- National Law Review: The Continued Proliferation of AI Exclusions (May 2026)
- Bloomberg Law: Insurer AI Exclusions Spark Policyholder Alarm on Coverage Gaps (July 2026)
- Cornerstone Research: Securities Class Action Filings Surge in the First Half of 2026 (July 29, 2026)
- Founder Shield: D&O Insurance Pricing, 2025 in Review, 2026 Outlook (January 2026)
- IndependentAgent.com: Verisk to Roll Out New General Liability Exclusions for Generative AI Exposures