The majority of carriers writing professional liability in Canada and the US have still said nothing about AI in their policy language. The minority that have said something are often saying it in ways that serve their own flexibility rather than their clients' protection.
That is the market a broker is navigating in 2026 - and the gap between what a policy appears to offer and what it will actually respond to at claim time has rarely been wider.
Natalie Chan, senior risk advisor, specialty, commercial insurance at Navacord, laid out the terrain at a recent Insurance Business Canada professional risk roundtable. "We are seeing three different approaches from insurers," she said. "Whether that be silent, affirmative, or exclusions. Explicitly inclusion, which means affirmative, does not necessarily mean better. Like all policy details, it's important to examine the actual insuring agreements, the definitions, the coverage trigger, or any other type of provisions that are contained in our policies."
Chan's three-category framework maps precisely onto what independent research is finding.
A 2026 RAND Corporation report, The Insurability of Artificial Intelligence, examined the market across public AI incident data, US litigation records, state legislation, and admitted-market insurance filings. It found the same three-way split: a minority of carriers affirmatively covering AI losses, a growing number filing broad exclusions, and the majority remaining silent. RAND identified the silent majority as the most immediately problematic category. When a policy says nothing about AI, coverage depends entirely on how a claim is characterized at the time of loss - and the carrier, not the broker or the client, controls that characterization.
The practical consequence is significant. A legal research tool that produces fabricated case citations could be framed as a professional error, a cyber incident, a product defect, or a general liability claim. Different policies respond to each framing differently, and none of them were written with AI specifically in mind.
The exclusion category is moving faster than most brokers have registered. Fenwick's June 2026 analysis found that coverage is fragmenting across cyber, tech E&O, D&O, and employment practices liability as insurers independently narrow AI protections within each line - what the firm described as "gap risk," where no single policy in a programme provides comprehensive coverage. That narrowing typically does not arrive as a single conspicuous AI exclusion. It arrives through revised base forms, tightening definitions, and restrictive carve-backs that a broker examining only the headline policy language will miss entirely.
Verisk's Insurance Services Office rolled out new general liability endorsements with January 2026 edition dates - forms CG 40 47 and CG 40 48 - that exclude AI-related bodily injury, property damage, or personal and advertising injury. Some carriers have moved further, filing what the market is calling absolute AI exclusions: broad bars on any claim arising from a company's use of AI, including AI-generated content and inadequate AI governance. These are general liability forms, but their significance for professional lines brokers is direct - when a claim crosses multiple coverage lines, a GL exclusion shapes how the overall programme responds and where the coverage dispute lands.
It is the affirmative category that carries the most dangerous misconception for brokers.
Alex Ilkos, client executive, professional services, at Purves Redmond Limited, described what he is actually seeing in the market at the same roundtable. "A lot of markets are putting wording in place, but I don't think a lot of coverage has necessarily changed on what's being offered," he said. "Things like affirmative AI coverage is more just there to say they're offering it, but I haven't seen a huge change in what's actually offered. So I think of it as a bit of a placeholder - that's preparing us for when they do have to make a change. It's a lot easier to change your endorsement wording to say, 'Hey, now we're excluded.'"
WTW's Insurance Marketplace Realities 2026 described the period between January 2025 and January 2026 as a "structural break" in the professional liability market - a sudden bifurcation in which some organizations renewed with affirmative AI coverage, albeit heavily conditioned, while others encountered sweeping exclusions. The key word is conditioned. Affirmative AI coverage that is heavily conditions-based - requiring documented governance, an AI risk assessment, and evidence of monitoring and oversight - is substantively different from affirmative coverage that simply names AI in the insuring agreement without conditioning it. A broker who cannot distinguish between the two is not in a position to tell a client whether they are actually covered.
Carriers that have moved into genuine affirmative AI coverage include Armilla AI with Lloyd's syndicate Chaucer, which expanded its offering in February 2026 under the name Vanguard AI, bundling cyber, technology E&O, and AI with explicitly named coverage for lawyer hallucination cases. Counterpart expanded affirmative AI coverage on miscellaneous professional liability in November 2025. Coalition offers AI coverage within its cyber program. These products exist - but they are not the same thing as an AI endorsement attached to a standard E&O form by a carrier that has not yet decided what it wants to do with AI risk.
For Canadian brokers, the professional liability AI conversation has a structural layer that does not exist in the same form in the US.
Individual lawyers in Canada must purchase their primary coverage through provincial law society plans - a government-mandated structure that removes much of the everyday broker from the immediate placement conversation. Ilkos flagged the consequence: "We're not seeing significant AI language in those provincial government-run policies yet. So we'll interestingly see how that will fall."
That wait has regulatory context behind it that is already moving. Legal Aid Ontario required roster lawyers to annually confirm compliance with the Law Society of Ontario's AI guidance starting January 2026. The Federation of Law Societies of Canada confirmed in 2025 that all 14 Canadian law societies endorse a single baseline posture: competence, confidentiality, and supervision duties apply to AI use without modification. The Law Society of BC's February 2026 amendments and the Barreau du Quebec's 2024 guidance add jurisdiction-specific requirements on top of that national baseline.
The regulatory accountability is already established. The insurance market has not caught up to it.
If provincial law societies move to exclude AI from their mandatory plans - a move that is being watched closely given the trajectory of the US regulatory environment - the question immediately becomes whether brokers can position drop-down excess coverage or an alternative primary policy to restore that gap. That is a product availability question as much as a placement question, and the market does not yet have a clear answer.
The underwriting picture adds a further dimension that brokers need to understand precisely.
Ilkos described what he is observing: "Of the AI governance-related questions we do see from the underwriting side, I'm personally not seeing huge impact on premium and terms offered. It's almost as if insurers are looking to gather this information, but they don't know yet what to do with it."
Chan agreed: "We're not seeing underwriters asking that many questions around the AI and the governance and the policies as of yet. Until we really see more claims, both in severity and frequency, as most of us can agree, our insurance industry does move slow."
The data on AI adoption in legal and professional services is running well ahead of that underwriting pace. According to the 2026 Legal Industry Report, 69% of legal professionals now report using generative AI tools for work-related purposes, up sharply from 31% just a year earlier. That is not a trend underwriters can ignore indefinitely - and the claims pipeline is building. Honigman's 2026 analysis found that one in five commercial insurers reported an AI-related loss in 2025, with only approximately half of those losses fully covered.
Chan's guidance to clients facing renewal in the current environment is direct: "Adopt a formal policy, or perhaps don't use AI. Because at that point, there are certain underwriters that want evidence that the organization is following, have rules, internal governance and guidance around the AI usage."
The broker who has already walked clients through that framework - and documented the conversation - is in a materially better position when a claim arrives and the coverage dispute begins. The broker who assumed the affirmative endorsement was sufficient is not.
