メインコンテンツへスキップ
Beancount.io LogoBeancount.io

ISO Generative AI Exclusions Start Jan 1, 2026: What CG 40 47 Means for Your CGL Policy

約3分Mike ThriftMike Thrift
ISO Generative AI Exclusions Start Jan 1, 2026: What CG 40 47 Means for Your CGL Policy

On January 1, 2026, Verisk's ISO introduced three optional generative AI exclusion endorsements for commercial general liability (CGL) policies: CG 40 47 (full exclusion), CG 40 48 (limited), and CG 35 08. Carriers are beginning to attach them, and Gallagher reports 47% of businesses plan to increase AI investment this year even as 38% already use AI for risk assessment — a coverage gap many small businesses have not noticed.

What CG 40 47 Excludes

CG 40 47 — Exclusion – Generative Artificial Intelligence (occurrence and claims-made CGL) — is the broad version. It excludes coverage under Coverage A (bodily injury/property damage) and Coverage B (personal and advertising injury) for loss arising out of generative AI, including outputs that contribute to bodily injury, property damage, or advertising injury. Language is broad enough that AI need only be a contributing factor.

Example from construction — an AI-assisted schedule or site plan that contributes to a defect claim — is exactly the scenario Gallagher flags as excluded even when AI was one tool among many.

CG 40 48 is a narrower variant; CG 35 08 addresses specific personal and advertising injury exposures. All three are optional at the insurer's discretion, but adoption is growing, and W. R. Berkley has gone further with an absolute AI exclusion across D&O, E&O, and fiduciary lines.

Why It Creates a Coverage Gap

AI-related claims do not fit neatly in one tower:

  • Bodily injury/property damage from an AI-generated instruction or design (CGL)
  • Advertising injury from AI-generated marketing copy that infringes or misrepresents (CGL Coverage B)
  • Professional failure from AI-assisted advice (E&O/PI)
  • Privacy/cyber from AI handling personal data
  • Management oversight of AI deployment (D&O)

If CG 40 47 sits on your CGL and a parallel AI exclusion arrives on your E&O or cyber tower, the same AI incident can be excluded in multiple places — a gap that did not exist when AI was treated as ordinary software.

What Businesses Should Do Before Renewal

  1. Ask for the endorsement schedule. At renewal, request a list of AI-related endorsements attached to CGL, E&O, cyber, and D&O. Do not assume "no AI exclusion" — confirm.
  2. Map AI use to risk towers. For each AI use — marketing, customer service, design, code generation — identify which policy would have responded before the exclusion and whether that tower now excludes.
  3. Consider standalone AI cover or sublimited affirmative grants. Some carriers now offer affirmative AI coverage riders; others place AI liability into cyber/E&O with AI-specific terms. Compare the cost of a rider versus the risk retained.
  4. Document human review. Policies that still cover AI-assisted work often require or reward a documented human-in-the-loop — verification of AI output before publication or deployment.

Bookkeeping Tie-In

Endorsements change the insurance asset you carry. Track premiums by tower and endorsement (Expenses:Insurance:CGL - Endorsement CG40.47) and maintain a coverage-map note in the ledger so the gap between premium paid and risk retained is visible at renewal, not after a claim denial.

Simplify Your Financial Management

Generative AI went from experimental to expected faster than insurance did. Beancount.io keeps AI-related costs, insurance premiums by tower, and coverage-gap reserves in version-controlled plain text — so the exclusion you didn't notice never becomes the claim you didn't cover. Get started for free and make AI a managed exposure, not an uninsured one.

この記事を共有