AI is exposing a blind spot in the insurance industry, with many insurers on the hook for risks embedded in policies that were never designed to cover increasingly autonomous AI systems. A report by the Artificial Intelligence Underwriting Company, co‑authored with researchers from Anthropic and Open AI, says that more than 90 per cent of insurers’ exposure to AI agents sits in “silent” cover embedded in conventional policies, leaving risks largely unpriced and often invisible to insurers themselves.

The review cites rising litigation as AI systems move beyond chatbots producing erroneous content to “agents” that take actions, exposing businesses to claims from professional negligence, cyber attacks and wrongful death. Notable cases include Google defending a lawsuit seeking at least $110mn after its AI Overviews system allegedly defamed US solar company Wolf River Electric, Air Canada being ordered to honour a discount invented by its customer service chatbot, and UK engineering group Arup losing HK$200mn after fraudsters used AI‑generated deepfakes of senior executives to trick an employee into transferring funds.

The authors warn that such cases foreshadow much larger liabilities as AI grows more autonomous, estimating a major AI disaster could cause about $100bn in direct damage with wider economic impact reaching trillions if insurers withdraw cover. They compare the risk to the terrorism insurance market collapse after September 11, when over $40bn of losses prompted insurer retreat. Experts such as Rajiv Dattani, Kevin Kalinich of Aon, and Matthew Botvinick of Anthropic urge dedicated AI coverage, new underwriting standards and technical auditing to keep pace with rapidly evolving models.