Evercore ISI expects more insurers will introduce policy language that specifically excludes claims caused by generated artificial intelligence (GenAI) to prevent silent coverage.
In a recent report, Evercore ISI noted that the growing use of GenAI and large language models (LLMs) is having an impact on insurance coverage, as users and others increasingly seek compensation for damages caused by their use.
Liability insurance for AI risks is attracting increasing attention given the large-scale lawsuits filed against model providers for financial losses and personal injuries due to LLM output errors, copyright infringement, and discrimination/bias.
Insurers are also paying attention Verisk’s Q4 2024 Emerging Issues Survey of Insurers found that a majority of respondents expect GenAI will lead to product liability claims within the next one to two years (32% expressed moderate concern and 23% expressed extreme concern).
As a result, insurers are introducing exclusions for claims arising from GenAI in policies to prevent silent coverage, while Verisk plans to introduce standard exclusion policy language in endorsements by 2026.
Evercore ISI said it was not aware of any insurance coverage for lawsuits involving GenAI model developers such as OpenAI or Anthropic, or companies using GenAI. However, the company believes these policyholders may assert third-party coverage under technical E&O, product liability, GL and EPLI policies, or first-party coverage under cyber and underlying property (business interruption). Evercore ISI believes technology E&O policies are most vulnerable to potential claims, although it believes the likelihood of this is currently low.
Evercore ISI said: “While it is early days and there do not appear to be any insurance claims related to negative LLM results, we are monitoring an emerging risk as AI adoption increases. The recent introduction of policy exclusions from Insurers/Verisk policy language indicates that insurers consider potential liability to be a possibility and are protecting themselves from silent coverage. We expect more insurers will introduce policy language to specifically exclude claims arising from genAI to protect against this risk even if a loss has not yet occurred (unlike silent cyber). Coverage in a GL/property policy that originally caused the loss before the policy language changed).
“We suspect large account insurers such as CB, AIG, Zurich, AXA​​
