The Silent Coverage Gap: Why Your Nonprofit's Insurance May Not Cover AI-Related Claims

"Silent AI" was the insurance equivalent of a gray area. Because AI tools were not mentioned anywhere in legacy policies, there was genuine ambiguity about whether AI-related incidents were covered. In practice, claims often proceeded under existing coverage frameworks, and insurers had limited ability to deny them solely on the grounds that AI was involved. The silence worked, imperfectly but reliably, in policyholders' favor.

The industry has been methodically closing this gray area. Carriers that initially tolerated silent AI coverage have been filing for explicit exclusions as they gain better insight into the frequency and severity of potential claims. The Verisk filing represents the industry's first coordinated, standardized approach to excluding generative AI from general liability policies at scale. When ISO files a new exclusion form, it is made available to all member carriers simultaneously, and most will incorporate it into their policy language at the next available renewal cycle.

The significance for nonprofits is timing. Organizations that renewed their general liability policies in early 2025 may still have the older, pre-exclusion language in place. But as policies come up for renewal throughout 2026, carriers are adding these endorsements. A nonprofit that deploys a beneficiary-facing AI chatbot today under a policy that was written before the Verisk forms may find that when that policy renews, the coverage it was counting on is simply no longer there.

Nonprofits are not merely at risk because they use AI. They are at heightened risk because of how they use AI. The populations nonprofits serve, including people experiencing homelessness, individuals in mental health crises, children, people with disabilities, and those navigating complex benefits systems, are among those most harmed when AI systems fail. The magnitude of potential claims is higher, and the moral stakes are more visible.

Healthcare insurers have become the most prominent example of AI-related litigation in analogous contexts. UnitedHealth Group faced a federal class action lawsuit alleging it used the nH Predict AI model to deny extended care claims for elderly Medicare Advantage patients. Cigna faced litigation for using its PXDX algorithm to automatically deny hundreds of thousands of claims with minimal human review. Both cases illustrate a principle that extends directly to nonprofits: when AI systems make decisions that harm people who rely on the organization's services, litigation follows, and the question of whether insurance responds becomes urgent.

For nonprofits, the scenarios where this risk materializes include intake AI that systematically deprioritizes certain populations based on biased training data; crisis chatbots that provide harmful advice to users in acute distress; eligibility determination systems that wrongly deny services; youth-serving programs where AI exposes minors to inappropriate content; and mental health organizations where AI-assisted assessments contribute to inadequate care recommendations. In each case, the potential for harm to vulnerable individuals is real, the potential for litigation is growing, and the question of insurance coverage is now deeply uncertain.

The good news is that the insurance market is beginning to respond to AI-specific risks with purpose-built coverage products. While these products are newer and in some cases more expensive than traditional insurance lines, they represent a genuine path to closing gaps that traditional policies are leaving open.

Armilla Insurance, underwritten by Chaucer Group through Lloyd's and launched in April 2025, offers an AI liability policy specifically designed to cover AI-specific perils including hallucinations, degrading model performance, and algorithmic failures. Relm Insurance has introduced PONTAAI, an excess difference-in-conditions wrap policy designed for organizations with third-party liability exposure from AI use, specifically designed to respond when existing liability programs exclude AI. These products represent the leading edge of a market that will grow significantly as AI incidents generate more claims data.

For most nonprofits, the immediate practical priority is not necessarily purchasing standalone AI insurance (though organizations with significant beneficiary-facing AI deployments should explore this). The priority is conducting an honest assessment of current coverage, identifying specific gaps, and having an informed conversation with your broker about options including affirmative AI endorsements on existing policies, increased limits on cyber coverage that does respond to some AI risks, and whether standalone coverage is warranted given your organization's AI use profile.

The connection between AI governance documentation and insurance is also important. Carriers writing AI liability coverage are beginning to use governance documentation as an underwriting factor: organizations that can demonstrate a documented AI governance framework, regular risk assessments, staff training on appropriate AI use, and defined oversight processes are more attractive risks and may qualify for better coverage terms. The NIST AI Risk Management Framework is a recognized standard that both insurers and regulators reference, and implementing it serves both risk management and insurability goals. This is explored further in our article on AI governance as risk mitigation.

The intersection of AI, insurance, and governance is most acute at the board level. Directors and officers have a fiduciary duty to ensure that organizational risk is understood and appropriately managed. When AI tools create coverage gaps that have not been identified, reviewed, or addressed by organizational leadership, that failure can itself become a source of liability, and one that the D&O policy may not cover if AI exclusions apply.

Harvard Law School's Forum on Corporate Governance identified this as a "hidden C-suite risk" in 2025, noting that AI failures can expose executives to derivative suits, regulatory enforcement, and reputational harm that their D&O policies may no longer cover. For nonprofit boards, the analogy applies with additional force: board members who authorized AI deployments without reviewing the coverage implications, and who did not require management to report on AI-related insurance gaps, may face questions about the adequacy of their oversight in the event of a significant incident.

The practical implication is that AI insurance coverage should be a standing agenda item at board risk committee meetings, not a topic that surfaces only when a claim is filed. Boards should require management to report annually on AI tools in use, the coverage position for each, and any changes in coverage terms at renewal. This is not an onerous requirement; it is the same level of oversight boards routinely apply to other organizational risks, and it is now equally important to apply to AI.

For more on building effective board oversight of AI risk, see our articles on D&O liability and AI for nonprofit boards and what boards need to know about AI insurance exclusions. For a broader view of AI risk management frameworks, our article on building an AI governance framework provides a practical starting point.