Insurers confront unpriced exposure from AI agents

More than 90% of insurers’ AI agent exposure may be embedded in conventional policies that were not designed for the technology, according to a report from the Artificial Intelligence Underwriting Company. The study found concentrated exposure in cyber, directors and officers, commercial general liability, and technology errors and omissions policies, where risks may be neither expressly included nor excluded.

Carriers are already adjusting policy language. Some have added affirmative AI wording to technology errors and omissions, professional liability and cyber policies, while a January 2026 ISO form allows exclusions for bodily injury, property damage and advertising injury from generative AI under standard CGL policies. Willis research found the professional liability market shifted between January 2025 and January 2026 toward explicit warranties or absolute exclusions.

AI agents raise distinct risks because they can carry out tasks, access company data and move funds with limited human oversight. Reported disputes and incidents include Arup’s HK$200 million (approximately $25 million) deepfake fraud loss in 2024, Wolf River Electric’s lawsuit against Google for at least $110 million, and a Canadian tribunal order requiring Air Canada to compensate a passenger misled by its chatbot.

The report warned that a severe AI event could produce around $100 billion in direct losses, with wider economic costs reaching into the trillions if insurers withdrew coverage. It called for dedicated AI cover, common technical standards and clearer policy language to reduce disputes over losses insurers may not have priced.