13 questions to ask your broker about AI exclusions.

"Is an AI exclusion endorsement already attached to any of our policies — and what are the form numbers?"

Why it matters: standardized AI exclusion endorsements (including ISO forms CG 40 47, CG 40 48, and CG 35 08) became effective from January 2026, and variants differ meaningfully in breadth. The form number tells you — and anyone reviewing your coverage later — exactly which language governs.

ISO endorsement filings, public state filing systems (SERFF)

"Will an AI endorsement attach at our next renewal — and when exactly is that renewal?"

Why it matters: exclusions generally attach at renewal, not mid-term. Your renewal date is therefore the deadline for everything else on this list — the documentation, the disclosure decisions, the market comparison. If you don't know the date, that's question zero.

"Which of our lines are affected — general liability only, or D&O, employment practices, and professional liability too?"

Why it matters: the exclusion wave started in general liability but is spreading into directors & officers, employment practices, and professional lines. A clean GL renewal can coexist with a newly excluded D&O policy — ask per line, not per company.

"Is the exclusion absolute — and how broad is the 'arising from' language?"

Why it matters: some carriers have filed absolute AI exclusions that disclaim coverage for anything "arising from" AI use — wording that can sweep in claims only loosely connected to an AI system. The breadth of that phrase determines what's actually at stake for your business.

Carrier form filings, public state filing systems (SERFF)

"Does any wording exclude claims tied to inadequate AI governance, policies, or training itself?"

Why it matters: some filed exclusions reach claims connected to the absence of AI governance — meaning your written policies and training records aren't just good practice; they can be the difference in whether a claim is even arguable. If this language is in play, documentation stops being optional paperwork.

"What AI use do we need to disclose on our application — and what counts as AI use?"

Why it matters: applications increasingly ask about AI, and staff often use tools nobody officially approved ("shadow AI"). Misstating use on an application creates its own coverage problems, independent of any exclusion. Agree with your broker on what counts before you sign anything. Expect a deflection here: most brokers won't take on defining the technical boundary of "AI use" — they aren't technologists, and the liability of drawing that line isn't theirs. The counter-move: ask for the carrier's own definition in writing, and agree on how this specific application's wording treats embedded features and assistants.

"If an employee causes a breach through an unsanctioned AI tool, does the AI exclusion apply — or does our cyber policy still respond?"

Why it matters: even with a strong acceptable-use policy and network controls, an employee will eventually paste sensitive data into an unapproved tool. Where that incident lands — denied under a new AI exclusion, or covered as an unauthorized act under existing cyber liability — is a boundary your specific forms decide, and it's exactly where coverage disputes will concentrate. Ask how your policies interact before there's a claim to argue about.

"How does AI embedded in our vendors' products affect our coverage?"

Why it matters: AI features now ship inside ordinary business software — office suites, CRMs, line-of-business apps. Whether a claim involving a vendor's embedded AI lands on your policy, the vendor's, or nobody's is exactly the kind of boundary question to ask while it's hypothetical. Expect a deflection here too: "review the vendor's license agreement" is the likely first answer, because third-party liability is genuinely messy. The counter-move: fine — then ask which of your policies would respond while that vendor dispute plays out.

"What documentation would strengthen our placement conversation?"

Why it matters: publicly reported underwriter interest keeps circling the same set: an inventory of AI systems in use, written AI policies, human-oversight procedures, training records, and incident-response plans. No outcome can be promised — but this is documentation you can present at renewal. One timing warning from people who sit in these meetings: this is the question to ask months before renewal, not at the placement conversation itself — by the time terms are on the table, the underwriter has already judged what you previously submitted. Asked early, this question is leverage; asked late, it's an admission.

"Are standalone AI liability markets relevant for our profile — and which ones?"

Why it matters: alongside the exclusions, an affirmative market has emerged — specialist programs including Armilla (Lloyd's coverholder), Munich Re's aiSure (via Mosaic Insurance), Corgi, Mayflower Specialty, and Testudo. Whether any fits your class, limits, and budget is precisely a licensed broker's question to answer.

Public company announcements, 2025–2026

"If a claim happens, how does timing work — and what about incidents from before the exclusion attaches?"

Why it matters: occurrence policies and claims-made policies treat the timeline differently, and an AI-related incident that predates the exclusion may be treated differently than one after it. The transition period around your renewal is where timing questions get expensive — ask them now.

"Can we see the actual endorsement text before renewal — not a summary?"

Why it matters: endorsement filings are public documents, and the exact wording is what governs a claim — not the summary, not the sales sheet. A broker can pull the text; you (or your counsel) should read the specific form headed for your policy.

Public rate & form filings, SERFF Filing Access

"What are underwriters asking about AI this year that they weren't asking last year?"

Why it matters: this question turns your broker into a forward-looking source. Whatever the answer is — new application questions, new documentation requests, new markets opening or closing — that list is your preparation plan for the next renewal, started a year early.