Verisk (formerly Insurance Services Office) and insurers have responded by drafting specific AI exclusions to limit their financial exposure. For insurance agents, brokers, underwriters and expert witnesses, simply reading liability policy language is no longer enough. To grasp the implications of these changes, you must understand the legal and historical context of such exclusions.

Understanding the critical details of AI insurance exclusions is essential for protecting your clients from devastating financial gaps. This article reviews the legal and historical forces shaping these emerging exclusions. We will examine how courts interpret similar policy language and analyze historical patterns in the insurance market. Through this comprehensive analysis, you will better recognize the escalating risk of potential coverage denials and implement strategic protection measures that better protect organizations’ interests.

The Significance of Insurance Contract Legal Language

In most U.S. states, judges consider insurance policies “contracts of adhesion,” meaning the insured must accept the policy as written or seek coverage elsewhere. Because insurers draft these contracts, courts have long recognized the unequal bargaining power between insurers and policyholders.

Courts have taken one of two approaches to this imbalance:

1. Contra proferentem
2. The reasonable expectations doctrine

Contra proferentem does not consider any extrinsic or parol evidence, such as oral or written communications between parties, when interpreting ambiguous terms. This law may be why underwriters are so often reluctant to answer coverage questions.

Courts that apply the reasonable expectations doctrine view the policyholder’s expectations of coverage much more broadly. Where there is ambiguity in the contract and courts cannot determine the “mutual intent” of both parties, the courts may consider what a reasonable person would have thought the word or phrase meant.[1]

As with any new exclusions, how courts interpret these exclusions will soon become case law. With generative AI exclusions, wording will become important in parsing cases. However, how practitioners use AI will matter, as well.

Will Minimal AI Involvement Trigger the Generative AI Exclusion?

Consider this example.

An insurance broker sells a commercial general liability policy with a generative AI exclusion to an association that uses generative AI to develop recreational outings for its annual meeting in Nashville. AI suggests the Grand Ole Opry as a place the meeting attendees “should not miss.” The insured then uses this recommendation for its attendees. However, during a nighttime outing to that venue, a mass casualty event occurs. Serious injuries occurred to two of the association’s attendees.

How will the courts interpret this use? Would the association’s reliance on generative AI to develop its agenda mean the carrier could deny this claim?

Some Insurers’ Broad Exclusions – Legal Context of Generative AI Coverage Exclusions

In response to threats associated with generative AI, the W.R. Berkley Corporation designed a new broadly worded “absolute” exclusion for use in the carrier’s D&O, E&O, and fiduciary liability policies. The exclusion applies to “…any actual or alleged use, deployment, or development of artificial intelligence by any person or entity…”[2] How courts will interpret this broad exclusion may hinge, according to one insurance blog, on the policy’s definition of “artificial intelligence.” This blog describes the development of this absolute exclusion as “an important development in how the insurance industry is navigating the complexities associated with AI.”[3]

The Three Most Litigated Words — “Arising Out Of”

While, to date, most of the generative AI exclusions focus on general liability policies, there is no doubt these exclusions will extend to professional lines of coverage. Employment practices liability insurance (EPLI), directors and officers coverage (D&O), and errors and omissions (E&O) coverage will see carriers, including advisory organization Verisk, add gen AI exclusions to these and other lines of coverage.

In agents’ E&O, Utica Insurance has already filed a gen AI exclusion; however, Swiss Re, another of the largest carriers offering insurance agents’ E&O, has not.

There are different “types” of exclusions, and the absolute AI exclusion favored by Berkley Companies is the broadest type. According to Harvard Law, the absolute exclusion will bar any claim “…based upon, attributable to, arising out of, or related to, in whole, or in part, any use of artificial intelligence, including but not limited to:

1. Errors, omissions, or inaccuracies in the programming, functioning, or the decision-making processes of artificial intelligence systems
2. Bodily injury, property damage, or economic loss caused by, or resulting from, the actions or decisions of artificial intelligence systems
3. Claims related to data breaches, privacy violations, or cyber incidents caused by the use of artificial intelligence systems.”[4]

Harvard Law reiterates in its article, “This exclusion applies to any party, including but not limited to non-insured parties, an insured, their employees, agents, contractors, or any third parties using artificial intelligence on the insured’s behalf.”

The article warns, and many professional liability experts agree, “The broad preamble to this exclusion, including coverage for claims ‘based upon, attributable to, arising out of, or related to, in whole, or in part,’ is incredibly broad and insurers may use them to exclude coverage for claims even where AI only played a negligible role in the wrongful act or error at issue in this claim.”

Clearly, agents should carefully read any exclusionary language, not simply assume the carrier uses ISO’s less restrictive exclusionary language.

Legal Context of Generative AI Coverage Exclusions Definitions Matter

When evaluating coverage under any policy that may include or exclude generative AI, agents must carefully review the policy’s definition of “artificial intelligence.” Agents should evaluate the following definitions if included.

• Deep machine learning
• Neural networks
• Chatbots
• Smart client portals
• Document completion software
• Self-learning systems

Finally, does your insured understand which automated systems fit the above definitions?

Clearly, definitions matter and the agent must understand how those definitions can impact the policyholder.

Contractual Shifts Are Needed to Hold AI Vendors Accountable

A Risk & Insurance article points out “…a structural imbalance in how liability is allocated between AI vendors and the enterprises that use their tools.” To date, courts and regulators view AI not as a potential codefendant, but simply as a tool. They expect that the user, the deploying organization, is responsible for any issues such as hallucinations or copyright infringement.[5]

Contracts with AI vendors favor the vendor, not the organization using the product. Contracts usually cap liability at one year of fees and do not guarantee the product’s performance.

New and targeted risk management measures surrounding contract negotiations and governance will be required, as further addressed in Part 3.

Insurers fear they may face “accumulation risk,” the potential for widespread losses from a single event—for example, a wave of intellectual property litigation.

Current Generative AI Litigation

According to Risk & Insurance, gen AI-related claims have risen by almost 1,000% since 2021. Furthermore, the report referenced in its article mapped “coverage gaps across every major policy line.”[6]

In one widely followed claim, class members succeeded in filing suit against UnitedHealth Group (UHG), alleging UHG allowed AI programs, rather than medical professionals, to deny claims in Medicare Advantage plans.[7] In addition, both Cigna Healthcare and Humana faced similar claims.

While these claims so far center on healthcare decisions, gen AI exclusions are so new that case law will continue to evolve. This inevitable evolution makes managing this risk difficult. Even with the best alternative insurance coverage, a November 2025 article by INSTECH experts recommends the following.

“…it matters less whether AI risks eventually sit within existing policy lines, emerge as embedded features, or evolve into a new, standalone class of AI insurance. What matters is that AI risks are material and growing, creating significant exposure to portfolios and businesses alike.”

The authors go on to say, “As such, they must be rigorously understood, quantified, and managed. Businesses adopting AI will need confidence that, when failures occur, clearly defined insurance coverage stands behind the technology if they decide to transfer the risk into the market.”