Last week, NPR reported that US insurers are spinning off standalone AI liability coverage as more than one in five American companies now run AI in daily operations. Deloitte projects a nearly $5 billion global AI insurance market by 2032. This matters to you because it signals what UK insurers will do next—and the gap between when your firm deploys AI and when your insurer can properly underwrite it is widening fast. The FCA's Consumer Duty PS22/9 and the PRA's SS1/23 both demand that regulated firms understand and manage AI risks explicitly. Your insurer will soon ask you to prove it. If you cannot, coverage will be denied, expensive, or simply unavailable.

This story is part of a larger pattern: the insurance industry is finally accepting that AI risk cannot be squeezed into traditional professional indemnity or cyber policies. It is a distinct liability class. The problem is not new—it has been obvious since 2023 when Harvey and LawGeex started automating contract review and legal firms began asking whether their E&O policies covered algorithmic error. What is new is that insurers are now admitting they do not understand AI risk well enough to price it into legacy products. They are right to be nervous. Unlike a software bug or a data breach, AI failure is often silent, distributed, and invisible until a client loses money or a regulator fines you. Lloyd's Blueprint Two and the EU AI Act both demand explainability and audit trails. Most UK firms deploying AI do not have them.

Trovix's view is uncompromising: buying AI liability insurance before you have proper AI governance is like buying a parachute after you have jumped. The insurers launching these products now—and yes, they will be good products—will require that you prove three things: first, that you know which decisions your AI systems make; second, that you can explain those decisions to a regulator or client; third, that you have audit trails and human oversight to catch failures before they cause harm. Luminance and other document AI vendors will tell you their products are 'explainable.' That is marketing. Explainability is not a feature of the tool—it is a consequence of governance. You need Trovix Audit or something equivalent to close the gap between deployment and understandability. Without it, your AI implementation is already uninsurable.

What should you do now? Stop. Audit your existing AI deployments—Copilot, contract automation, client intake, compliance screening, whatever you are running. Document what each system does, who validates its output, and how you would explain a failure to the FCA or SRA. This is not optional post-launch work. It is the price of admission to the next round of AI liability insurance. Then, before you deploy new AI systems, build governance into the plan from day one. Trovix Watch can flag changes in AI regulation in real time so you do not fall behind. Your insurer is about to demand this evidence. Your regulator already does. The firms that move now will have cheap, broad coverage. The firms that wait will have none.

Source: NPR Marketplace