AI risks make some insurers wary of corporate liability

Corporate insurance policies are beginning to exclude coverage for AI-related harms, making deployments of generative AI potentially riskier for businesses.

Insurers are increasingly hesitant to provide corporate liability coverage for Artificial Intelligence (AI) risks due to emerging harms and a lack of standardized data for pricing these new liabilities. A report by the RAND Corporation highlights that AI-related harms, such as incorrect outputs, deepfakes, privacy violations, and discriminatory decisions, create a demand for insurance that does not fit existing lines. Companies are seeking protection from the financial repercussions of deploying unreliable AI, but insurance providers are wary of the unknown risks. This wariness is leading some insurers, like W. R. Berkley, to explicitly exclude AI-related incidents from Directors and Officers (D&O), Errors and Omissions (E&O), and Fiduciary Liability policies. Similarly, Verisk/ISO has introduced optional language for carriers to exclude generative AI harms from property and casualty policies, impacting over 80 percent of US policies. While some new and existing companies are attempting to fill these coverage gaps, the industry faces challenges in developing a common taxonomy for AI incidents and claims, with RAND advocating for clearer regulations and notice requirements for AI coverage.

The insurance industry's caution around AI risks signals a potential slowdown in enterprise AI adoption, as companies may delay projects to ensure they can adequately cover fiduciary obligations. This hesitancy stems from the difficulty insurers face in quantifying and pricing the novel risks associated with AI, such as misinformation, deepfakes, and autonomous failures, which do not neatly fit into existing insurance frameworks. The lack of standardized data and clear regulatory guidance exacerbates this problem, leading to policy exclusions and coverage gaps.

For businesses actively deploying AI, the increasing likelihood of exclusions in liability policies means they must proactively manage AI risks or potentially self-insure against them. This situation could drive demand for specialized AI risk assessment tools and services, and may prompt a push for clearer industry standards and regulatory oversight to establish predictable coverage. As RAND suggests, the development of a common taxonomy for AI incidents and claims is crucial for the market to mature and for AI risks to become more insurable.