Claims checked
Two claims made in a LinkedIn post of 15 September 2026, both resting on secondary sources:
- ISO's January 2026 CG 40 47 endorsement lets insurers exclude certain generative-AI bodily injury and property damage from commercial general liability.
- More than 80% of AI exclusion requests from major carriers were approved by state regulators by April (Wolfe Research analysis of state filings).
Public reporting
CSIS, Gregory C. Allen, "The Insurance Industry's Retreat from AI Threatens to Slow Innovation and Adoption", 4 September 2026
- The 80 percent figure: "On April 23, 2026, The Information reported that state insurance commissioners had quietly approved more than 80 percent of carrier requests to exclude AI-related damages from corporate insurance policies." CSIS attributes the figure to The Information, not to Wolfe Research.
- Wolfe Research appears in CSIS with a different number: "approximately 800 consumer AI-related lawsuits were filed against U.S. businesses in 2025, up 140 percent year-on-year."
- "the ISO Generative AI Exclusion for commercial general liability took effect on January 1, 2026."
- Carriers filing AI exclusions: Berkshire Hathaway, Chubb, Travelers, AIG, Tokio Marine Holdings, W.R. Berkley, Great American, Fairfax Financial. "more than 60 property and casualty insurance providers in 2026 filed for AI exclusions".
- W.R. Berkley's proposed endorsement covers "any actual or alleged use" of AI (first reported by the Financial Times, November 2025).
- "On July 10 ... Verisk's Insurance Services Office confirmed it was drafting coverage exclusions for agentic AI."
- Source: CSIS, The Insurance Industry's Retreat from AI Threatens to Slow Innovation and Adoption
Claims Journal, Don Jergler, 20 July 2026
Quotes Verisk's descriptions of the three forms: CG 40 47 (bodily injury, property damage, and personal and advertising injury arising out of generative AI), CG 40 48 (personal and advertising injury) and CG 35 08 (bodily injury and property damage, products-completed operations). No adoption figures.
Sources that could not be read
- A Verisk article on core.verisk.com (no readable content).
- The Insurer, 10 July 2026 (paywall): Verisk weighs new exclusions for agentic AI risks.
- The Information, 23 April 2026 (paywall; not attempted): Berkshire Hathaway, Chubb Win Approval to Drop AI Insurance Coverage.
Form text
ISO CG 40 48 01 26, Exclusion: Generative Artificial Intelligence (Coverage B Only)
Read in full, from a copy hosted by ALM Media: CG 40 48 01 26 (PDF). One page, self-identifying as "CG 40 48 01 26 © Insurance Services Office, Inc., 2025".
- It adds to the Coverage B exclusions: "This insurance does not apply to: 'Personal and advertising injury' arising out of 'generative artificial intelligence'."
- Definition: "'Generative artificial intelligence' means a machine-based learning system or model that is trained on data with the ability to create content or responses, including but not limited to text, images, audio, video or code."
ISO CG 40 47 01 26 (Coverage A and B) and CG 35 08 (products-completed operations)
Form text not read. No public specimen was found. The scope of all three forms is reported by Claims Journal (20 July 2026, quoting Verisk), and the scope of CG 40 47 also by Gallagher.
State filings
State filings in SERFF were not read. SERFF Filing Access requires accepting its terms of use.
Agentic AI
Verisk's statement, as reported: "we are evaluating additional options to address AI-related exposures, including agentic AI" (The Insurer, 10 July 2026, paywalled). CSIS (4 September 2026) says ISO "confirmed it was drafting coverage exclusions for agentic AI". Not read at Verisk.
Findings
- Confirmed from the form text: ISO has a generative AI exclusion for commercial general liability personal and advertising injury, edition 01 26, with the definition above (CG 40 48, read in full).
- Not confirmed: the exact wording of CG 40 47; "more than 80 percent approved" (attributed to The Information by CSIS, and to Wolfe Research by the LinkedIn post and some newsletters; neither original source read); "60+ insurance groups filed" (The Insurer, 23 July 2026, paywalled, repeated by CSIS); the agentic AI exclusion (reported, not read at Verisk).
- Attribution: CSIS credits the 80 percent figure to The Information (23 April 2026), and gives Wolfe Research a different figure (about 800 consumer AI lawsuits in 2025, up 140 percent).
