Opening Chapter
Volume XII:
The Hardening.
The E&S market does not harden quietly. It hardens the way glaciers move — imperceptibly until the moment a serac collapses and rewrites the valley below. We are in that moment now. Admitted carriers are shedding property books in coastal California and hurricane corridors as if the paper itself were combustible. Cyber underwriters who wrote anything with a pulse in 2021 are now demanding network segmentation audits, incident response retainers, and a detailed account of why your client's CFO still clicks links in phishing emails. Professional indemnity lines are watching the AI liability question metastasize in real time, with no agreed case law, no consensus exclusion language, and a claims bar that smells blood.
What the actuarial models cannot capture — what no loss-development triangle fully encodes — is the texture of practitioner knowledge: the adjuster who has seen this exact fact pattern before, the agent who watched a wholesale broker decline the same risk three consecutive years before a specialty market finally priced it right, the underwriter who knows the difference between a coverage gap and a coverage trap. That knowledge lives in conversations. This forum is where those conversations happen, with the rigor they deserve and the candor that only peers afford.
When the Admitted Market Exits, the E&S Market Does Not Simply Fill the Void.
The withdrawal of admitted carriers from high-hazard coastal and wildland-urban interface risks is not a temporary dislocation. It is a structural re-pricing of the social contract between insurer and insured — one that the E&S market is being asked to absorb at velocity it was not designed to handle. The surplus lines mechanism exists precisely for this purpose, but the assumption has always been that admitted carriers maintain a floor. That floor is rising.
Independent agents working Florida residential accounts in 2025 are not placing the same risk they placed in 2019. The property has not changed. The exposure has not meaningfully changed. What has changed is the carrier's willingness to price for tail risk at any premium the insured can bear. The conversation has shifted from "what does this cost?" to "can this be placed at all?"
I had a $4.2M coastal frame in Pinellas County. Filed with seven admitted carriers. Seven declinations. The E&S market came in at 3.4× the prior year premium and still wanted a $50K hurricane deductible. The client renewed. What choice did they have?
The underwriting discipline required in this environment is not simply risk selection — it is risk translation. The E&S underwriter who can articulate to the admitted market why a risk is actually more favorable than the zip code suggests, who can structure the coverage to make the exposure legible to a reinsurance treaty, who can write the endorsement language that closes the gap without creating a new one — that underwriter is the scarcest resource in the market right now.
Admitted Carrier Exits — Coastal Property, 2020–2025
Insured cat losses, US market, 2024 · Third consecutive year above $100B
Increase in ransomware claims frequency, 2020–2024
Average total cost of a data breach, US, 2025
Policyholders with cyber coverage who have never read their exclusions
Average dwell time before detection in manufacturing sector incidents
The Cyber Policy Is Not the Product. The Coverage Language Is the Product.
The single most consequential professional failure an agent can commit in 2026 is placing a cyber policy without reading the war exclusion. Not the war exclusion in the abstract — the specific language in the specific policy form, because they are not the same across carriers and the difference is not academic. The Merck v. ACE litigation established that "hostile or warlike action" is interpretively ambiguous in the context of state-sponsored cyberattacks. Carriers responded not by clarifying the exclusion, but by expanding it, in ways that vary by form, by endorsement, and by manuscript language negotiated at placement.
The agents in this forum have been having this conversation for three years. They know which carrier forms contain the LMA5567 endorsement. They know which wholesale brokers will negotiate the Lloyd's cyber war exclusion versus which ones will simply bind what the market offers. They know the difference between a silent cyber exposure in a CGL policy and an affirmative cyber grant in a stand-alone form — and they know how to explain that difference to a CFO who thinks his technology errors & omissions policy covers a ransomware event.
From the Forum — Cyber Working Group
The carrier just denied the claim citing the infrastructure exclusion. The attack vector was a third-party SaaS vendor. Their argument is that the insured's own systems weren't compromised. We're in coverage litigation now. Anyone else seen this exact fact pattern?
I pulled the endorsement language from six carriers this week. LMA5567 is in all of them, but the definition of 'critical infrastructure' is different in every one. This is not a market that has standardized its war exclusion. It has standardized the name of the exclusion.
Editor's Annotations
Peer Access
These conversations continue — with 340 more threads this quarter alone.
Forum members receive full thread access, the quarterly brief, and invitations to vertical working groups.
"The AI liability question is not a future problem. It is a present problem wearing a future problem's clothes. The claims are already being filed. The coverage language was written for a different world."
AI-Assisted Professional Services: The Coverage Gap No One Agreed to Create
The standard professional liability policy was designed for a world in which the professional made the decision and the tool executed it. A lawyer researches the case. A doctor diagnoses the patient. An architect designs the structure. The professional's judgment is the insured act. The tool's output is the artifact of that judgment.
Generative AI inverts this relationship. The tool produces the output — the contract clause, the diagnostic suggestion, the structural analysis — and the professional reviews it. The question that no policy form has cleanly answered is: when the professional relies on that output and it is wrong, whose professional judgment failed?
The claims bar has not waited for the insurance market to answer this question. Suits are being filed against law firms for AI-assisted research errors, against accounting firms for AI-assisted audit anomaly detection failures, against medical groups for AI-assisted diagnostic support errors. The defendants' E&O carriers are reading their professional services definitions and finding that "professional services" was defined in 2018, by a drafter who was not thinking about large language models.
The underwriters in this forum are doing what practitioners do in the absence of regulatory clarity: they are developing informal standards through conversation, testing coverage language against hypothetical fact patterns, flagging the exclusions that will not hold under judicial scrutiny, and building the institutional knowledge that will eventually become the market's consensus position.
Increase in AI-related professional liability claims, 2024 vs 2023
Standardized AI liability exclusion forms currently adopted across admitted E&O carriers
"We are writing policies for risks we don't fully understand, in a legal environment that doesn't yet exist. That's what surplus lines is for. But even we need better language."
— E&S Underwriter, Midwest Regional · Forum Thread, Jan 2026
Has anyone successfully negotiated a manuscript endorsement that affirmatively covers AI-assisted professional services without carving out the specific tool? The carriers I've approached want to name the system in the endorsement. That's going to be obsolete in 18 months.
We wrote a manuscript form for a large law firm last quarter that defines the insured act as "the professional's exercise of judgment in reviewing, approving, or declining AI-generated output." It's held up through one renewal. Whether it holds up through a claim is a different question.
We are seating you at the right table. Tell us who you are.
The forum is not open enrollment. It is curated by practice area, so that a claims adjuster untangling a complex liability dispute is in the room with underwriters who write that line — not with agents who have never seen the back end of a claim.
Membership is granted on the basis of professional context, not credentials. We are not gatekeeping. We are seating.
What Membership Includes
Full access to all forum threads, indexed by vertical and coverage question
Invitations to vertical working groups (Property Cat, Cyber, PI/E&O)
Quarterly brief — 12,000 words of practitioner analysis, formatted for print
Direct introductions to peers in your practice area
Governs under Chatham House Rule — attribution only with consent