Generative AI Exclusion (CG 40 47 / CG 40 48 / CG 35 08)

A set of Verisk-filed Commercial General Liability endorsements removing coverage for bodily injury, property damage, and personal and advertising injury arising from generative AI.

The Generative AI exclusion refers to a set of standard endorsements filed by Verisk (which owns ISO and publishes the standard policy forms it issues) and attaching at U.S. Commercial General Liability renewals from January 1, 2026. Endorsement CG 40 47 is the broad form on Coverage A and Coverage B, removing bodily injury, property damage, and personal and advertising injury arising out of generative AI. CG 40 48 is the Coverage B only variant, removing personal and advertising injury alone. CG 35 08 is the Products and Completed Operations Liability variant, removing bodily injury and property damage arising out of generative AI inside that coverage part. Together they convert silent AI on the underlying CGL form into an explicit exclusion.

The endorsements are the industry's collective answer to the silent AI problem on Commercial General Liability. CGL was never designed to cover financial-loss claims arising from AI outputs, but the silent wording invited litigation over edge cases (an AI-driven decision that caused physical harm, a chatbot-generated personal injury). The new exclusions remove the ambiguity by stating clearly that generative AI claims are not covered on the standard CGL form, pushing the exposure into either an affirmative endorsement or a standalone Generative AI Liability policy.

Carriers have broad latitude in how they adopt the endorsements. Some attach CG 40 47 by default at every renewal; others adopt CG 35 08 to preserve narrow coverage; others draft proprietary AI exclusions that go further or further-narrower than the Verisk wording. The result is that no two CGL policies will treat AI identically in the post-January 1, 2026 environment, and the buyer must read the actual endorsement schedule, not assume Verisk standard wording.

Several Tech E&O, Cyber, and Professional Liability carriers have followed the CGL move with their own AI-specific exclusions, either tracking the Verisk language or drafting proprietary equivalents. The combined effect on a deployer's program is that affirmative AI coverage is now usually required to fill the gap; without it, a Gen AI claim can fall into the space between excluded forms with no policy designed to answer.

Also known as

Verisk Generative AI Exclusion, ISO Generative AI Exclusion, CG 40 47, CG 40 48, CG 35 08

Frequently asked

How do CG 40 47, CG 40 48, and CG 35 08 differ?

CG 40 47 is the broad form on Coverage A and Coverage B of the standard CGL: it removes bodily injury, property damage, and personal and advertising injury arising out of generative AI. CG 40 48 is the Coverage B only variant, removing personal and advertising injury but leaving Coverage A intact. CG 35 08 is the Products and Completed Operations Liability variant, removing bodily injury and property damage arising out of generative AI inside that coverage part. Carriers select among them based on appetite and which coverage part they want to neutralize.

When did the generative AI exclusions take effect?

Verisk filed the endorsements for attaching at U.S. Commercial General Liability renewals from January 1, 2026. Individual carrier adoption varies (some attached the wording at every renewal from that date forward; others adopted on a state-by-state schedule as filings cleared local insurance departments). The effect is rolling rather than instantaneous: a CGL policy issued in late 2025 may not yet carry the endorsement; the same insured's renewal in 2026 almost certainly will.

Do the exclusions affect Cyber and Tech E&O policies?

Not directly; CG 40 47 and CG 40 48 are standard endorsements on the Commercial General Liability Coverage Part, and CG 35 08 on the Products and Completed Operations Liability Coverage Part, not on Cyber or Tech E&O wordings. However, several Cyber and Tech E&O carriers have filed their own proprietary AI exclusions, often tracking the Verisk language as a model. The practical result for a deployer is similar: AI exposures that were silently inside the older form are explicitly outside the new form, and the gap has to be filled with affirmative coverage.

Related terms

generative AI liability insurance overview

General information, not legal or insurance advice.