AI and professional indemnity insurance: what your broker may ask

A market-wide survey of Lloyd's underwriters, published in mid-2025, asked which classes of business would carry the heaviest AI-related losses. Professional indemnity came out top, ahead of cyber, product recall, and accident and health. For a UK accountancy, law, consultancy or surveying practice, that finding turns AI and professional indemnity insurance from a background worry into a live renewal question.

The Lloyd's Market Association survey heard from 144 managing agents (the firms that run Lloyd's insurance syndicates), 94% of them underwriters. Their reasoning was consistent: professional services already use AI widely, and the losses could be sizeable where erroneous AI-assisted advice or analysis causes harm to a client. For a 12-person practice, the practical shift is this: you now need to be able to describe how you use AI, not just use it well.

AI and professional indemnity insurance: why insurers are naming it in the wording

This is not a hypothetical shift. From June 2026, CFC, a specialist commercial insurance provider in the London market, began adding affirmative AI wording across a run of policy lines: technology errors and omissions, professional liability, eHealth, intellectual property, management liability, media and cyber. It completed that rollout by extending affirmative AI cover to its media policy from 30 July 2026.

"Affirmative" cover means the policy language addresses AI explicitly, rather than leaving it to be argued over after something has already gone wrong. Nick Line, CFC's chief underwriting officer, put the shift plainly: "AI is no longer an emerging technology for media companies. It is already embedded in the way many businesses create, manage and distribute content." Swap "media companies" for accountancy, law or consultancy and the sentence still holds. Full details of the June 2026 rollout are here, and the July extension is covered by Insurance Business. One insurer writing explicit AI wording does not mean every insurer has. But it shows underwriters are no longer treating AI use as background noise.

What happens when AI-assisted work goes wrong

The Lloyd's survey was about where the exposure sits. A recent High Court case shows what one of those losses looks like in practice. In R (Ayinde) v Haringey LBC, the Divisional Court found that generative AI tools had been used to produce written legal arguments and witness statements that were never checked against real sources. The result was fake citations and quoted passages that did not exist in the material they claimed to come from.

The court's warning was direct: freely available generative AI tools built on large language models are not capable of conducting reliable legal research, and often produce answers that are "apparently coherent and plausible" but "entirely incorrect", citing sources that do not exist or quoting passages that are not really there. Read the case summary here. That is precisely the shape of loss the Lloyd's survey had in mind: erroneous AI-assisted work reaching a client, or a court, unchecked.

What this means for a small practice

Being able to explain how AI is used in client work, and where a person checks the output before it leaves the building, is becoming part of the renewal conversation. It is worth preparing for rather than being caught out by.

That is a different bar from simply having a sensible AI policy. It means being able to answer questions on the spot, in a call with your broker, without scrambling for an example the night before.

The questions to be ready for at renewal

There is no standard AI questionnaire here, and your broker may ask nothing like this. We have drawn these from where the survey says underwriters' concern sits. They are the questions a professional practice should be able to answer about its own AI use. Go through them before the call, not during it.

1. Which parts of client work involve AI: drafting, research, analysis, summarising, or client communication?
2. Which tools are formally approved for that work, and which ones have staff picked up on their own initiative?
3. Where in the process does a person check AI output before it reaches a client, and is that checkpoint written down anywhere, or does it just happen in someone's head?
4. What is meant to happen if an AI tool gets something wrong: who is meant to catch it, and when was that last actually tested?
5. Could you produce a written record of your AI use today, or would it take a week to put together from scratch?
6. Has anyone in the practice used a generative AI tool for legal, technical or financial research without checking the output against a primary source?
7. Who owns AI use in the practice: is there a named person accountable for it, or is it nobody's specific job?

Shadow AI is what catches most practices out: tools staff have picked up on their own that nobody higher up knows about. If a partner cannot answer question two with confidence, that is usually why. Our piece on shadow AI in small businesses covers how to find out what is actually in use before a broker asks.

Question four is the Ayinde risk in miniature: an AI tool producing something that sounds right and is not, with no human catching it before it goes out. We have written separately about why AI agents can be confidently wrong, and why human sign-off matters even when the tool has performed well every time before.

For accountancy practices specifically, where AI is now doing meaningful work on drafting and analysis, our guide to building an AI workflow accountants can stand behind is worth reading alongside this one.

Building the record before you need it

Answering these questions well starts with having a written account of your own AI use, not a mental one. We have set out a practical way to build that: our guide to running a same-day AI governance audit walks through the framework for producing a one-page record you can point to. That post covers building the record. This one is about who is now asking to see it: your broker, at renewal.

None of this is insurance advice, and nothing in this article tells you what your policy does or does not cover. Ask your broker directly, and check the wording of your own policy rather than assuming.

If you are not yet sure where AI touches client work across your practice, the free HoursBack quiz is a plain-English starting point. Take the quiz and get a clear list of where AI is being used well, and where nobody has checked.