The speed of the exclusion wave caught most risk managers off guard. One carrier updated their E&O policy. Then two more. Then Verisk/ISO issued a standardized CGL form with an AI exclusion, which means it will filter into the renewal terms of policies across the entire market. By the time most legal and risk teams noticed, significant coverage had already quietly disappeared.
This piece maps what each major exclusion actually removes, in plain terms, so you can assess the gap in your current program before you need to make a claim.
The exclusion map: carrier by carrier
| Carrier / Issuer | Policy line | What the exclusion removes | Effective |
|---|---|---|---|
| Berkley | D&O, E&O | "Absolute" AI exclusion. Removes coverage for any claim where AI was involved in the decision chain. This includes AI-assisted decisions, not only fully autonomous ones. | Mid-2024 |
| Philadelphia | Tech E&O, Cyber | AI-generated output exclusion. Removes coverage for losses arising from content, recommendations, or decisions produced by an AI system, including third-party models your product calls via API. | Late 2024 |
| Hamilton | E&O, Cyber | Autonomous action exclusion. Specifically targets agentic systems. Any loss from an AI acting without real-time human approval falls outside coverage. | Early 2025 |
| Verisk / ISO | CGL (standardized form) | Broad AI exclusion. Removes bodily injury, property damage, and personal and advertising injury arising from AI use. Because this is an ISO form, it propagates into the renewal language of most commercial GL policies at next renewal. | January 2026 |
"The Verisk/ISO form is the one that changes the calculus for everyone. It does not require a carrier to make an active decision; it simply rolls in at renewal. Most policyholders will not notice until they file a claim."
What "AI-assisted" actually means in a claims context
The Berkley exclusion's reference to "AI-assisted decisions" is the broadest and most consequential. If your team uses an AI tool to surface a recommendation, and then a human approves it, Berkley's position is that AI was in the decision chain. That means a D&O claim arising from that decision may not be covered.
This is not hypothetical. We have seen early-stage disputes where carriers are arguing that because a risk model flagged a credit decision, or an AI tool drafted a contract clause that was then signed, the AI-assisted exclusion applies. The question of what qualifies as "in the decision chain" will be litigated for years.
Which lines still have appetite for AI risk
Not every carrier is retreating. The appetite exists; it has moved to specialists. Lloyd's syndicates with technology expertise are actively underwriting AI risk. A small number of MGAs, including Quark, have built policy forms specifically for AI exposure. The coverage exists; it is just no longer embedded in standard renewal packages.
The key distinction underwriters are drawing: were you negligent in how you deployed the AI, or did the AI fail despite reasonable safeguards? Companies with documented risk assessments, monitoring in place, and human oversight at key decision points are getting very different terms than companies that deployed without any of that infrastructure.
What to do before your next renewal
- Pull your current E&O, Cyber, CGL, and D&O policy forms and search for "artificial intelligence," "machine learning," "automated decision," and "AI." Note which policies have added exclusions since your last renewal.
- Map which of your products or workflows use AI. Categorize by whether AI outputs are advisory (human approves) or autonomous (AI acts directly).
- Identify which customer contracts include AI liability indemnification clauses. This is where your first claim is most likely to originate.
- Get an underwriting view from a carrier or MGA that is actively writing AI risk, before your renewal rather than after an incident.
The exclusion wave is not finished. We expect additional carriers to update CGL and umbrella forms through 2026 as the ISO form propagates. The window to replace coverage proactively, before a claim tests your current program, is narrowing.
Not sure what your current policy covers?
We'll review your existing program and identify the gaps. No obligation.