The exclusions created the demand.
A first-party line for AI execution loss, structured for the surplus-lines market.
THE GAP
Nothing your client holds covers it.
Cyber
Built for a hacker getting in. Not for software the client authorized that executed wrong.
Errors and omissions
Pays a claim made against them. Nobody is claiming.
Crime
Requires a dishonest person. There isn't one — the AI agent is not a person, and its credentials were valid.
A loss like this looks like liability, so it gets filed under a policy never built to respond—and declined on grounds nobody saw coming.
THE MARKET
The exclusions are attaching.
Attaching at renewal, silently.
Verisk's ISO AI endorsements carry a January 2026 edition date as carriers attach them at discretion. Your client won't notice—it only shows on the schedule. Reading it now puts you ahead of the loss instead of explaining it later.
Exclusion is how the market writes.
Verisk notes exclusions give underwriters stability to keep writing risk rather than walk away. While rational for them, it shifts the exposure directly onto your client—and the hard questions back onto you as their trusted broker.
This is how cyber started early.
Dozens of P&C groups have filed to adopt AI exclusions. Producers who secured early cyber books hold them a decade later. This is new premium on accounts you already service—and an opportunity to lead before someone else does.
WHAT WE CAN WRITE
Three questions decide if a risk is bindable.
01
Can they see what it did?
Every action is visible end to end—whether built in-house, run on a leading agent platform, or inherited inside a vendor product.
02
Are the logs tamper-evident?
Kept in a record the agent cannot reach or amend, and retained securely for a defined minimum period without exception.
03
Controls against expensive mistakes?
Permissions encoded before production. Change control tracked on every update, with a named individual explicitly accountable for each workflow.
Three yeses and it's bindable. Two and we'll tell you which one to fix.
THE TERMS
The shape of the terms.
Basis
First party. Surplus lines.
Trigger
Discovery-based Deviation Event. One root cause, one deductible, one limit.
Rated on
Registered workflows, permissions, controls. Not revenue.
Limits
Indicative limits: $1M per event standard; up to $5M on referral. Aggregate is a multiple of the per-event limit.
Terms
Retention and notice terms are returned with the indication on submission.
BRING US A RISK
Tell us what the client's AI does.
We will quickly let you know if it is a risk we can write. Submissions are straightforward: share what the AI does, its permitted scope, and the surrounding controls. Indications are generated from the questionnaire, with record verification occurring prior to binding. Specimen wordings are available under an NDA upon indication of interest.
