Sometimes, but don't assume it. Most policies don't mention AI, so an AI-caused mistake is treated like any other: professional indemnity may cover negligent advice, general liability covers injury or damage, cyber covers data breaches. The gaps are chatbot promises, copyright in AI images and new AI exclusions. Get written confirmation from your broker.
This is a practical map, not legal advice. Policy wordings differ, and only your broker or insurer can confirm how yours responds. What you can do yourself, in about an hour, is list where AI touches your customers and your work, match each use to the policy that would respond, search your documents for AI wording, and send your broker the specific questions at the end of this tutorial.
Match each AI mistake to the policy that would respond
Insurance follows the type of harm, not the tool that caused it. An error typed by a tired employee and an error produced by an AI assistant land on the same policy, unless the policy says otherwise. So start with what could go wrong.
| What goes wrong | Example | Policy most likely to respond | Where cover can fail |
|---|---|---|---|
| Wrong advice or a flawed piece of work for a client | An AI-drafted report contains a wrong figure the client relies on | Professional indemnity, also called errors and omissions | AI exclusions in the wording; services you never declared to the insurer |
| Your chatbot promises something you don't offer | It tells a customer they can return sale items after 90 days | Often none. The cost is honouring your own promise, not damages | Most liability policies don't pay for refunds, discounts or contract terms you agreed to |
| Injury or damage caused by AI-written content | Wrong allergen or care information in a product description | General liability or product liability | New optional exclusions for injury and damage "arising out of" generative AI |
| Copyright, trade mark or defamation in AI content | A marketing image closely resembles another designer's work | The personal and advertising injury section of general liability, or a media liability policy | Intellectual property is often only partly covered already; AI-specific exclusions now target this section |
| Customer data leaks through an AI tool | A staff member pastes a customer list into a free chatbot | Cyber | Proposal-form answers about your controls; conditions on security practices |
| You lose money directly | An automation sends the wrong price to every customer and you honour it | Usually nothing, unless a cyber or crime policy has a relevant first-party section | Liability policies cover claims made against you by others, not your own losses |
Two rows cause most surprises. The chatbot promise feels like an insurable error, but paying out what you said you'd pay is usually treated as a business cost. And direct losses, where nobody sues you and you simply lose money, sit outside most liability cover whether or not AI was involved.
The AI exclusions now arriving at renewal
For years, AI risk was "silent": policies neither covered nor excluded it by name, so a claim was judged on the general wording. That's changing, and the change often arrives quietly in a renewal pack.
- Standard general liability endorsements. Verisk, whose ISO policy forms many insurers use as the base for their general liability wording, released three optional generative AI endorsements with a January 2026 edition date. CG 40 47 removes cover for bodily injury, property damage and personal and advertising injury arising out of generative AI. CG 40 48 removes only the personal and advertising injury part. CG 35 08 does the same for products and completed operations cover. They're optional, so they only apply if your insurer adds them.
- Broad exclusions on management and professional policies. In 2025, the insurer Berkley introduced what it called an "absolute" AI exclusion for its directors and officers, errors and omissions, and fiduciary liability products. It excludes claims based on "any actual or alleged use, deployment, or development of Artificial Intelligence", and lists chatbot promises and AI-generated content among its examples.
- Wide definitions. Verisk defines generative AI as a system "trained on data with the ability to create content or responses, including but not limited to text, images, audio, video or code". That catches the AI built into ordinary office software, not only the chat assistant you chose to buy.
Search every policy document for these words: artificial intelligence, generative, machine learning, algorithm, automated and chatbot. Check the endorsements and schedules as well as the main wording, because exclusions are usually added as endorsements. If you find one, don't panic and don't cancel anything. Ask your broker what it removes and whether it can be narrowed or taken off.
Five things that decide whether a claim is paid
- The exact wording, including endorsements. Two policies with the same name can respond very differently. The endorsement page matters as much as the headline cover.
- What you told the insurer. Proposal forms increasingly ask about AI use. An answer that was accurate when you gave it can become wrong within months as your team adopts new tools. If you added a customer-facing chatbot mid-year, your broker should know. Picture an illustrative tutoring company that answered "no" to "Do you use AI in providing services to clients?" at last year's renewal. Six months later its tutors started using an assistant to turn session notes into termly progress reports for parents. One report mixed two pupils' notes and gave a parent another child's mock exam results. When the owner rang the broker about a possible complaint, the first two questions were whether AI had been involved and whether that use had been declared. How much that matters depends on the policy's disclosure conditions, but it's a conversation you want to have before the incident, not during it. A six-monthly reminder to re-read your proposal answers costs nothing.
- Whose loss it is. A claim by a customer (third party) and a cost you bear yourself (first party) are handled by different parts of your insurance, and many small businesses only buy the third-party parts.
- Whether it was a promise or a mistake. In a 2024 tribunal case, an airline argued that its website chatbot, which had given a customer wrong information about a bereavement fare refund, was "a separate legal entity that is responsible for its own actions". The tribunal rejected that, found the airline liable for negligent misrepresentation and ordered it to pay the customer a few hundred dollars. You own what your chatbot says, and the bill for honouring it is often yours alone.
- Whether you took reasonable care. Negligence claims turn on whether you acted sensibly. Evidence that a person reviewed AI output, that the chatbot was tested and limited to safe topics, and that staff had rules to follow helps both your defence and your insurer's view of you.
Three illustrative claims, three different cover outcomes
These are illustrations of how the same question plays out differently, not predictions of how any insurer would decide.
An online clothing shop's chatbot invents a returns policy
Say the shop's website chatbot tells a customer that sale items can be returned within 90 days. The real policy is 14 days and excludes sale items. The customer posts the screenshot, and over a week thirty more customers ask for the same deal. The shop decides honouring it is cheaper than the complaints: perhaps $1,800 in refunds on stock it can't resell at full price. Insurance is unlikely to help, because nobody has claimed damages; the shop is paying for its own promise. The fix is operational: restrict the chatbot to quoting the returns page word for word and handing anything else to a person.
A wine merchant's AI tasting notes make a health claim
Say a wine merchant uses AI to write product descriptions for 200 new wines, and one describes a wine as vegan when it was fined with an animal product. A customer with a relevant allergy reacts and makes a claim. This is bodily injury from a product, so general or product liability would normally be the policy to respond. The question is whether the renewal added a generative AI exclusion to that cover. The fix: a person checks every health, allergen and dietary statement against the supplier's specification sheet before it goes live.
A handmade jewellery seller's AI campaign image
Say a jewellery seller generates lifestyle images for a seasonal campaign, and one closely resembles another designer's signature piece. A letter arrives from the designer's lawyer. Some general liability policies cover limited intellectual property claims arising from advertising, which is exactly the section the CG 40 48-type endorsement removes for generative AI. Many small sellers have no media liability cover at all. The fix: photograph your own products, use AI only for backgrounds or mock-ups you check, and keep a record of how each image was made.
Specialist AI insurance: who it's actually for
A small number of insurers now write cover that names AI risk explicitly. Munich Re says it has offered AI insurance since 2018; its aiSure products are aimed at AI vendors backing performance warranties and at corporations deploying models at scale, and cover areas such as contractual liabilities, hallucinations, intellectual property and financial losses from AI failures. Armilla offers affirmative AI performance and liability insurance for businesses deploying AI across their operations.
For a small business using off-the-shelf tools, specialist cover is rarely the first move. Better value usually comes from closing gaps in the policies you already have: asking for AI exclusions to be removed or narrowed, getting written confirmation of how each policy treats AI-assisted work, and keeping AI away from decisions where a mistake is expensive. Look at specialist cover if you sell a product or service built on AI, or if a client's contract requires you to carry it.
What to send your broker before renewal
Brokers answer specific questions far better than "are we covered for AI?". Send something like this at least a month before renewal, and keep the reply with your policy documents.
Subject: AI use - questions before our renewal
We use AI in these ways:
1. [e.g. drafting client reports, always reviewed by a person]
2. [e.g. a website chatbot answering delivery and returns questions]
3. [e.g. AI-generated images in social media advertising]
For each policy we hold (general liability, professional indemnity,
cyber, and any others), please confirm in writing:
a) Does the current wording, or any renewal endorsement, exclude or
limit cover for claims arising from artificial intelligence or
generative AI? If so, please send the exact wording.
b) Would a claim arising from each of the uses above be treated the
same way as the same mistake made by an employee?
c) Do we need to declare any of these uses, and is our proposal
form still accurate?
d) Can any AI exclusion be removed or narrowed, and at what cost?
e) Is there a gap you would recommend we close, and what would it
cost to close it?Filled in by an illustrative five-person marketing agency, the top of that email reads:
We use AI in these ways:
1. Drafting website copy and social captions for clients. An account
manager edits and approves every piece before the client sees it.
2. Generating campaign images for client adverts with an AI image
tool, about 30 a month. Final images are approved by the client.
3. Summarising client calls with a meeting-notes tool. Summaries are
emailed to the client after each call.Say the broker's reply (illustrative) says the professional indemnity wording has no AI exclusion, but the general liability renewal adds an endorsement removing personal and advertising injury "arising out of" generative AI. That answer is useful, and it isn't finished. Use 2 is the risky one, because an image that resembles someone else's work is an intellectual property claim, and the endorsement has just removed the section that might have answered it. The follow-up questions write themselves: does the professional indemnity policy cover intellectual property claims arising from work done for clients, and what do the agency's client contracts say about who carries that risk? Use 3 is worth one more line too, because a call summary sent to a client that gets a price or deadline wrong is advice the client may rely on.
Make the risk smaller, not only the policy bigger
Insurers price what they can't see as high risk. Most of what reduces your exposure also makes you easier to insure.
- Keep a list of every AI use, who owns it and what could go wrong. A simple AI risk register takes an afternoon and gives your broker exactly what they need. Two rows from an illustrative six-person estate agency show the level of detail that helps:
| AI use | Owner | What could go wrong | Control | Policy that would respond | Broker confirmed? |
|---|---|---|---|---|---|
| Assistant drafts property descriptions from viewing notes | Senior negotiator | States "south-facing garden" or a tenure the property doesn't have; a buyer relies on it | Every factual line checked against the property fact sheet before listing | Professional indemnity | Yes, in writing, 3 March |
| Website chat answers viewing and opening-hours questions | Office manager | Promises a reduced fee or a viewing slot that doesn't exist | Limited to FAQ page answers; anything about fees goes to a person | Probably none for a promise; the agency bears the cost | Asked, awaiting reply |
- Put a person between AI and anything that advises, prices or makes a health or safety claim. Setting up human review doesn't have to slow the work down if you target it at the risky outputs.
- Limit what customer-facing AI can say. A chatbot that quotes policy pages and hands everything else to a person can't invent a refund. See who is liable when a chatbot gets it wrong for how to scope one safely.
- Read your AI vendor's limitation of liability clause. Business terms usually cap what the vendor will pay, often at the fees you paid over a recent period, so the vendor won't cover your loss.
- Have a plan for the day it goes wrong. Knowing who switches the tool off, who contacts affected customers and when you notify your insurer is covered in an AI incident response plan. Late notification can itself put a claim at risk, so check the notification clause.
If your worry is specifically data leaks and cyber incidents, the questions insurers ask there are different; what cyber insurers ask about AI covers them.
A decision rule you can use today
If AI only helps staff draft internal work that a person checks before anyone relies on it, your current cover probably treats it like any human error, but confirm that once in writing. If AI talks to customers, sets or quotes prices, or produces content you publish, get written confirmation for each policy, search for AI exclusions at every renewal, and tighten the controls above before you add more AI rather than after.
Follow-up questions about AI and your insurance
Do I have to tell my insurer that my staff use ChatGPT?
It depends on the questions in your proposal form and the disclosure conditions in your policy, which differ between insurers. Many renewal forms now ask about AI directly, and an answer that was true last year may not be true now. If AI talks to your customers, sets prices or produces work clients rely on, tell your broker in writing and keep their reply.
Does a disclaimer on my website chatbot protect me?
Only partly. A notice that answers may be wrong sets expectations, but in the 2024 airline chatbot case the tribunal held the business responsible for information on its own website, chatbot included. A disclaimer is unlikely to undo a specific promise the bot made. Limiting what the chatbot is allowed to answer protects you far more than wording at the bottom of the window.
Will the AI vendor pay if its tool causes the mistake?
Rarely. Business terms usually make you responsible for checking outputs and cap the vendor's liability, often at the fees you paid over a recent period. Some business plans add a narrow promise to defend copyright claims about outputs; Microsoft's Customer Copyright Commitment is one, with conditions such as keeping the built-in content filters switched on. Read your own plan's terms.
Further reads
- What Are the Risks of Using AI in My Small Business? — The wider list of AI risks, beyond the ones insurance touches.
- How to Mention AI Use in Client Contracts and Proposals — How to word AI use in contracts so promises match your cover.
- How to Write an AI Usage Policy for Your Small Business — A written policy is often the first thing a broker asks to see.
- AI Hallucinations Explained for Business Owners: Causes and Fixes — Why AI invents facts, and the checks that catch it.
- What to Check in an AI Tool's Privacy Policy and Terms — Read the liability and data clauses in your AI tools' terms.
- AI Compliance Checklist for Small Businesses: What Applies to You — The rules that apply to your AI use, alongside insurance.
- What an AI Consultant Can't Do for You, and What You Must Own — Seven responsibilities that stay with the business when you hire AI help, an ownership card for every automation, and promises no consultant should make.
- What to Do When AI Gets Something Wrong With a Customer — The first hour after an AI error reaches a customer: whether to honour what it said, apology wording for email and phone, and how to stop it happening again.
- AI Readiness Checklist for Accountants, Solicitors, Consultants — Twenty checks, grouped and scored, that tell an accountancy, law or consulting firm whether it is ready to pilot AI or has gaps to fix first.
- What Does AI Cost a Small Insurance Brokerage in Year One? — Year-one AI costs for four, eight and twelve-person brokerages, quarter by quarter, including the staff time and compliance work most budgets leave out.
- How Brokers Use AI to Compare Policy Wordings at Renewal — A two-pass method for renewal wordings: a text compare that can't skip a page, then AI to explain each change, graded on a shared rubric and checked at source.
- AI Acceptable Use Policy for a Small Professional Firm — A clause-by-clause checklist and one-page sample policy for accountants, lawyers and consultants who handle confidential client work.
- Where AI Saves Time in a Small Insurance Brokerage — One commercial renewal followed step by step, showing where AI cuts the minutes, where it adds errors-and-omissions risk, and which jobs to automate first.
- How to Write a Health and Safety Policy With AI and What to Check — Draft a three-part health and safety policy with AI from your real hazards, then check it on a walk-round before anyone signs it.
- Business Continuity Planning With AI: A Template for Small Firms — A ten-section continuity plan template with the checks for each item, plus prompts that let AI interview you and run a tabletop test.
- Compliance Calendar: Track Licences, Insurance and Certificates — Track every licence, policy and certificate in one register, let AI pull the dates from new documents, and send reminders early enough to act on.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Verisk ISO generative AI exclusion forms CG 40 47, CG 40 48 and CG 35 08 (summary published on IndependentAgent.com); Hunton Andrews Kurth insurance recovery commentary on Berkley's AI exclusion; Munich Re aiSure product pages; Armilla product information; published summaries of the 2024 airline chatbot tribunal decision; Microsoft Customer Copyright Commitment documentation.