A manufacturer I spoke to last month had been running AI-driven quality inspection across their packaging lines for two years.

In February, a batch of defective components made it through. They filed a claim.

Three weeks later, the insurer responded. Their January 2026 renewal had added an AI exclusion to the policy on page 47 of a 63-page document. 

Their broker had never raised it and the risk manager had never read that far.

The claim was denied in full.

Before we get into what changed and when, quick question:

👋🏻 I'm Leonardo Ubbiali. This week we're looking at why the insurance coverage most manufacturers assume they have for AI deployments quietly disappeared in January 2026, and what it takes to get it back.

The manufacturer I spoke to is not unusual.

For years, AI claims landed inside standard commercial policies because nothing excluded them. 

Insurers called this "silent AI."

The same pattern ran through cyber risk from 2015 to 2023. Manufacturers discovered ransomware attacks were excluded only after the claim arrived.

The Insurance Services Office introduced Form CG 40 47 on January 1, 2026.

It underpins roughly 82% of global property and casualty policies. Verisk confirmed strong carrier interest before the forms were even published.

The exclusion appeared in endorsement schedules and renewal questionnaires. By the time it was standard, most policyholders were already bound by it.

The manufacturing industry is about to learn this the same way it learned silent cyber — one denied claim at a time.

Take a machine-vision system inspecting parts on a packaging line.

It misses a defective component. The line kept moving. The part shipped. The customer found it three weeks later.

A recall follows.

CG 40 47 excludes bodily injury, property damage, and personal and advertising injury arising out of generative AI. If the AI made the call that let the defect through, the standard CGL policy does not respond.

Tech E&O policies cover AI developers. Not the manufacturers deploying vendor models. Your vendor's contract almost certainly disclaims responsibility for model outputs. That liability lands on you.

The coverage that used to absorb it is gone.

The pattern is documented beyond the manufacturer I spoke to.

A mid-market firm running AI-powered customer service workflows had a chatbot generate an unauthorized commitment to a client. 

The client relied on it and then filed a suit. 

The firm's E&O policy, renewed in January 2026, had the same exclusion buried in the endorsement schedule. The claim was denied in full.

Legal commentators are calling it the same trajectory as silent cyber, a coverage gap that grows quietly until a claim forces the issue.

The manufacturers getting ahead of this pulled their renewals, found the exclusions, and started building what underwriters now require before they will quote: model registries, human oversight logs, and incident reporting protocols.

Munich Re's aiSure, Armilla, and the handful of tech E&O extensions filling the gap all require the same three things. None will quote without them.

Most of those manufacturers built the governance infrastructure ahead of the EU AI Act's high-risk AI deadline, originally set for August 2026, though the timeline remains in flux.

The insurability came as a side effect. The ones without it are facing both problems at once, and the clock is running on both.

A prioritized list of coverage gaps by deployment type, the governance documentation checklist underwriters require, and the questions your broker should be able to answer before you sign.

Covers the legal implications of the new ISO endorsements, how courts may interpret prior policy years, and what policyholders should be doing at renewal.

Time to value: 12 minutes

Does your risk manager know which AI systems in your plant are no longer covered going into Q2 renewals?

Hit reply.

I read every email.
Leo