The speed of AI development has made its risk largely unpriced and invisible, and it is a time bomb for insurers, a new report says.

The bulk of insurers’ exposure sits as silent coverage inside existing policies such as cyber, professional and general liability, according to San Francisco-based Artificial Intelligence Underwriting Company, building on research by AI companies Anthropic and Open AI and insurance companies including QBE and Aon.

AI risk is increasing with the development of agent models that can receive instructions, form plans and then take actions with limited or no human oversight.

“The AI agent economy, projected to be handling trillions of dollars’ worth of transactions by 2030, is the next such development whose risks insurers must price, manage and absorb,” the report says.

Exposure is growing rapidly and insurers risk repeating the “costly ambiguity that plagued cyber insurance”.

The tech’s capability is outpacing its reliability, and because it is dominated by only a few models – such as Open AI’s ChatGPT, Anthropic’s Claude and Google’s Gemini – there is a risk of correlated losses.

The report suggests insurance industry players collaborate to build shared infrastructure encompassing incident data collection, catastrophe modelling, standards, contract design, risk selection, pricing, monitoring and claims management.

It says organisations that use AI will need insurance for first-party operational losses, where an agent damages the policyholder’s own systems, data, operations or reputation, and for legal liability, where an AI agent as product or service harms a business partner, customer or member of the public.