As generative artificial intelligence becomes more integrated into commercial operations, insurers are adjusting policy language to address emerging risks. At the same time, litigation tied to AI development, deployment, and disclosure continues to grow, prompting broader use of exclusions and endorsements in liability policies.
Insurers have introduced provisions designed to limit or eliminate coverage for AI-related exposures. However, the scope and enforceability of these exclusions remain uncertain. Some carriers require narrower coverage through endorsements, while others offer policies with affirmative AI-specific coverage. Courts have yet to fully address how these exclusions will apply, particularly where coverage could be significantly limited.
AI-Related Claims Span Multiple Legal Theories
Recent litigation reflects a wide range of AI-related claims. These include copyright and intellectual property disputes tied to training large language models, such as Bartz v. Anthropic, which reportedly settled for $1.5 billion. Product liability and negligence claims have also emerged, including Raine v. OpenAI, Inc., involving alleged real-world harm.
Privacy and data-use claims continue to develop, as seen in Reddit, Inc. v. Anthropic PBC, which challenges the use of scraped data for AI training. Antitrust allegations, such as Chegg, Inc. v. Google LLC, focus on the use of proprietary data in AI systems. Additional cases involve discrimination and algorithmic bias, including Mobley v. Workday, Inc., as well as securities class actions tied to disclosures about AI capabilities, such as D’Agostino v. Innodata Inc.
This breadth of litigation has led insurers to adopt broader exclusionary language.
Insurers Introduce Broad AI Exclusion Language
Some carriers have implemented what they describe as “absolute” AI exclusions in directors and officers, errors and omissions, and fiduciary liability policies. These provisions may apply to claims arising from AI use, related disclosures, regulatory requirements, or alleged violations of AI-related laws.
In some cases, definitions of artificial intelligence are expansive. One insurer defines AI as any machine-based system that generates outputs such as predictions, content, or decisions based on input data. This definition can include tools that have been used in business operations for years.
Other insurers have adopted targeted exclusions focused on generative AI, including systems that produce text, images, audio, or synthetic data in response to user prompts. These provisions may reference tools such as ChatGPT, Bard, Midjourney, or DALL-E.
The Insurance Services Office has also introduced optional endorsements for commercial general liability policies. These include exclusions applying to bodily injury, property damage, personal and advertising injury, and products and completed operations liability when claims arise from generative AI.
Courts May Interpret Exclusions Narrowly
Legal standards governing policy interpretation remain relevant. Courts generally construe coverage provisions broadly and interpret exclusions narrowly. Insurers may still have a duty to defend claims that include both AI-related and non-AI-related allegations.
Coverage may also apply when an alleged injury could occur independently of AI use or when claims do not clearly fall within the exclusion. In cases involving mixed allegations, insurers may be required to defend the entire action if any portion is potentially covered.
Courts have also addressed illusory coverage, holding that exclusions should not eliminate the fundamental protections a policy is intended to provide. Additionally, earlier policies issued before AI-specific exclusions may still offer coverage, as changes in policy language over time can influence how courts interpret prior forms.
Stay informed and ahead of the curve — explore more industry insights and program opportunities at ProgramBusiness.com.