# Does Business Insurance Cover Mistakes Made by AI?

Canonical URL: https://business.theaieducator.io/posts/business-insurance-cover-ai-mistakes
Publication: Dan Fitzpatrick Insights
Author: Dan Fitzpatrick
Topic: AI Governance and Risk
Published: 2026-11-08T16:40:04.000Z
Modified: 2026-11-08T16:40:08.417Z

Since January 2026, US insurers can exclude generative AI losses from standard policies, while some London insurers now cover them. Here is what the evidence shows, a one-page AI Use Map, and the questions to put to your broker.

## In brief

Business insurance may not cover AI mistakes. Since January 2026, US insurers can attach standard endorsements that exclude generative AI losses, while some London insurers now cover AI affirmatively. Map every AI use to a policy, then get your broker's answer in writing with form numbers.

## Key takeaways

- Since January 2026, ISO has offered three optional generative AI exclusions for commercial general liability policies: CG 40 47, CG 40 48 and CG 35 08.
- Trades Coverage counted 4,078 filed exclusion adoptions across 49 US states and DC with data through 31 July 2026, but filings do not show what sits on an issued policy.
- UK treatment differs for now: CFC added affirmative AI wording across seven products in July 2026, and the Lloyd's Market Association was waiting for a market instruction on exclusions.
- Different AI mistakes land on different policies: general liability, errors and omissions, cyber and management liability can each be triggered.
- The Claim Path Check asks of every AI use: which policy would you phone first, and does its current wording still say yes?
- The AI Use Map turns that question into one table you can send to your broker, with a column for the answer in writing.
- Ask for form numbers and written answers; a verbal reassurance at renewal is a hope, not cover.

Maybe, and it depends on a few lines of your policy that may have changed this year. Since January 2026, US insurers have been able to attach standard endorsements that remove cover for losses arising from generative AI. Whether yours did is written in your renewal documents, and your broker can confirm it in writing. Check before a claim, not after.

That short answer hides a split. Many owners assume the policy they bought before ChatGPT quietly covers everything their staff now do with it. Some policies do. Some now say they do not. The rest of this article shows which words to look for, how the US and UK markets differ, and a one-page map you can fill in this week and send to your broker.

## Do standard business policies still cover AI mistakes?

Not automatically, and the wording is moving fast. In January 2026, Verisk's ISO, the organization that writes the standard policy forms many US insurers copy, made three optional generative AI exclusions available for commercial general liability policies. Verisk [announced the forms](https://www.independentagent.com/vu_resource/verisk-to-roll-out-new-general-liability-exclusions-for-generative-ai-exposures/) with a January 2026 edition date, and the Big "I" independent agents' association carried the notice.

The three forms work like this, according to [Insurance Business](https://www.insurancebusinessmag.com/us/news/professional-liability/ai-exclusions-are-spreading-through-commercial-policies-brokers-warn-clients-to-check-cover-594489.aspx) (2026, undated on the page I read):

- **CG 40 47** removes cover for bodily injury, property damage, and personal and advertising injury arising out of generative AI.
- **CG 40 48** applies the exclusion to personal and advertising injury only.
- **CG 35 08** applies it to products and completed operations cover.

"Optional" matters. An insurer has to choose to use a form, and then attach it to your policy. Gallagher, a large broker, [describes the CG 40 47 endorsement](https://www.ajg.com/news-and-insights/iso-introduces-generative-ai-exclusion-in-commercial-general-liability-policies/) (July 2026) as one that is "beginning to appear" on commercial general liability policies and may create coverage gaps that "aren't immediately obvious."

## How many insurers have actually adopted AI exclusions?

A lot of them, though the count measures paperwork and not your policy. Trades Coverage, which publishes insurance guidance for US trades, searched state insurance filings and [reported](https://tradescoverage.com/reports/genai-exclusion-adoption/) (August 2026, data through 31 July 2026) 4,078 filed form adoptions across 49 US states and the District of Columbia, up from zero in mid-2025. Each adoption is one carrier or program filing one generative AI exclusion for use in one state. Of the 4,078, 3,955 involve coverage lines that contractors and small businesses commonly hold, such as general liability, umbrella, businessowners and errors and omissions.

Read that with care, as the report does. The filings do not show whether a carrier attached an exclusion to an issued policy, and the forms are optional unless a carrier's program rules require them. My reading: the direction is clear and the effect on your policy is not. The only way to close that gap is to ask.

The same report says there are currently no admitted standalone generative AI products to buy back what the exclusions remove. That is the report's finding about the US admitted market on its data date, and I have not been able to verify it independently.

## Is the UK different?

For now, yes, though not by much and not for long. A July 2026 [report from Resultsense](https://www.resultsense.com/news/2026-07-30-cfc-affirmative-ai-cover-lloyds-market/) says the specialist London insurer CFC added affirmative AI wording to its media policy from 30 July, completing a program across seven products. Under that wording AI involvement does not stop the policy responding. The same report says ISO, AIG and Berkley filed exclusions in the US, and that the Lloyd's Market Association was waiting for an instruction from the market before drafting equivalent exclusions.

So two policies at similar levels of cover may answer the same AI mistake in opposite ways. A [DAC Beachcroft presentation](https://lmalloyds.com/wp-content/uploads/2026/04/AI-and-Policy-Wordings-DACB-21-April-2026.pdf) to the Lloyd's Market Association wordings forum (21 April 2026) shows how broad some proposed AI exclusion language can get, reaching claims about content created with AI, failure to detect AI content, and even a chatbot's alleged promises. I treat it as a sample of what could appear, not as a wording in force.

I did not find a UK survey of how many small business policies carry an AI exclusion today. Treat any figure you are quoted for the UK with suspicion until you can see its source.

## Which policy answers when AI gets it wrong?

It depends on what went wrong, and that is why one exclusion is not the whole story. Bruce Gendelman Insurance told Insurance Business that AI incidents could trigger general liability, errors and omissions, cyber and management liability cover, depending on the circumstances. Its examples were a chatbot giving customers incorrect information, an AI-drafted contract containing errors, and AI-generated marketing that infringes someone's rights.

Put plainly: a wrong estimate that leads to a wall failing is a different claim from a wrong paragraph in an advice letter, and each lands on a different policy. Trades Coverage gives a trade example. A contractor might use AI to prepare an estimate or diagnose equipment, then do the physical work by hand, and an exclusion could give the carrier a basis to deny a later claim that the AI-assisted step contributed to. Whether it would is a question for the policy wording and the facts, not for me.

This is general information, not insurance or legal advice. Your broker and, for a dispute, a solicitor or attorney have to read your actual documents.

## What is the Claim Path Check?

The Claim Path Check is one question asked of every way your business uses AI: if this went wrong and a customer sent you a bill, which policy would you phone first, and does its current wording still say yes? A use of AI you cannot answer for is an unmapped exposure, whatever else you have done well.

I coined it for this article as a suggested way of thinking, not as established research. It has three steps:

1. **List the uses.** Every place AI touches a customer, a contract, a quote, advice, marketing or money.
2. **Trace the path.** For each, name the policy that would respond and the person who would phone the insurer.
3. **Ask for the words.** Ask your broker whether the policy carries a generative AI exclusion or an affirmative AI wording, and get the answer in writing with the form number.

## What does the AI Use Map look like?

Copy this table into a document and fill in one row per use. The first row is a worked example, using a hypothetical 14-person heating and plumbing firm, not a real business.

| AI use | What could go wrong | Policy I think responds | Question for my broker | Answer in writing (yes, no, unclear) |
|---|---|---|---|---|
| Drafting quotes and job estimates (hypothetical) | A wrong figure or missed safety step is relied on | General liability, possibly professional cover | Does any generative AI exclusion apply to work that began with an AI-drafted estimate? | |
| Website chatbot answering customers | It promises a callout or warranty we do not offer | | | |
| Drafting adverts and social posts | Copied wording or a false claim | | | |
| Summarizing contracts or tenders | A clause is missed or invented | | | |
| Staff pasting client data into a chatbot | Confidential data leaves the building | Cyber | | |

Two columns carry the weight. "Policy I think responds" shows you where your own assumptions are thin. "Answer in writing" is the only column an insurer will be held to. A verbal reassurance from a busy renewal call is a hope.

## What I Tell Business Owners

The first question owners ask about AI risk is usually about the law or the tool. The one I would ask first is about the policy. I have given more than 500 keynotes in 30 countries, and across the business teams I work with, AI use is almost always wider than the owner thinks and wider than any insurance conversation has covered.

That is why I keep returning to a governing principle: outsource the doing, not the thinking. An assistant can draft the quote. A person has to decide who answers if the quote is wrong, and that includes who pays. The Loop Test applies here too. Before you accept "a human checks it" as your safeguard, ask whether that person would notice a wrong result without being told to look, in time to stop it mattering, and whether the work could be put back if they missed it. I would expect a claim to probe the same ground, though that is my inference, not a finding.

## Which response to the exclusion gap holds up?

There are four common responses. I compared them against the sources above on 8 November 2026.

| Response | What the sources say | Fits when | Breaks when |
|---|---|---|---|
| **1. Assume you are covered** | Filings are rising fast, but they do not show what is on your policy (Trades Coverage). Broker advice is to check, not assume (Insurance Business). | Never as a plan | You find out at the claim |
| **2. Map your uses, then get written answers** | Mapping where AI is used is the first step brokers recommend. Affirmative and exclusionary wordings now differ across the market (Resultsense). | Nearly every business that uses AI with customers | Your broker will not put anything in writing |
| **3. Stop using AI in the exposed jobs** | An exclusion can bite where AI touched one step of a larger job (Trades Coverage). | The job is high consequence and the check is weak | It makes your team hide AI use rather than stop it (see shadow AI) |
| **4. Ask about affirmative or specialist cover** | Some London insurers now write AI cover affirmatively (Resultsense). A specialist AI liability product was announced for Lloyd's (Verisk notice). I found no admitted standalone US product (Trades Coverage). | After step 2, when the map shows a gap that matters | It is used instead of fixing how the work is checked |

The sensible order is 2, then 3 or 4 for the gaps that remain. Response 1 is the one most businesses are on by default.

## What should you do this week?

1. **Monday, 30 minutes.** Fill in the first two columns of the AI Use Map. Include uses by individual staff, not only approved tools. Staff often use AI nobody approved, which is the [shadow AI](https://business.theaieducator.io/posts/shadow-ai-small-business) problem, so ask them.
2. **Tuesday, 15 minutes.** Find your renewal date and the name of the person at your broker who handles it. Pull the policy schedule and look for any form number starting with CG 40 47, CG 40 48 or CG 35 08, or any wording that mentions "generative," "artificial intelligence" or "chatbot."
3. **Wednesday.** Email the broker the map and ask three things: does any policy carry an AI exclusion, does any carry affirmative AI wording, and which policy would respond to each row. Ask for the form numbers.
4. **Thursday, one hour.** For every row marked "no" or "unclear," decide: stop, add a real check, or ask about different cover. Write one owner's name beside each decision.
5. **Friday.** Put a date in the diary to repeat the check at every renewal and whenever a new AI use goes live. Add the map to your [AI policy](https://business.theaieducator.io/posts/does-a-small-business-need-an-ai-policy).

If you must tell customers when AI is involved, [AI disclosure for small businesses](https://business.theaieducator.io/posts/ai-disclosure-small-business) covers what to say. If your team is checking AI output, [how to train staff to check AI output](https://business.theaieducator.io/posts/train-staff-to-check-ai-output) is the practical follow-up.

## Where to go next

If you run a trade or construction firm, [AI for trades and construction](https://business.theaieducator.io/posts/ai-for-trades-and-construction) shows where the exposure tends to sit. For customer-facing chat, see [AI for customer service](https://business.theaieducator.io/posts/ai-for-customer-service-small-business), and when a supplier says it uses AI on your behalf, my [supplier questions](https://business.theaieducator.io/posts/questions-to-ask-suppliers-who-use-ai) include one about who carries the risk. For the board-level version, see [board questions about AI](https://business.theaieducator.io/posts/board-questions-about-ai-small-business).

If you want to know where your business stands before you spend another month experimenting, take the [Business AI Readiness Scorecard](https://theaieducator.io/business-ai-readiness?utm_source=business.theaieducator.io&utm_medium=referral&utm_campaign=business-insurance-cover-ai-mistakes). If your team needs to learn to check, correct and use these tools well, that is what my [AI training for teams](https://theaieducator.io/ai-training-for-business?utm_source=business.theaieducator.io&utm_medium=referral&utm_campaign=business-insurance-cover-ai-mistakes) is built for. And if you are planning an event for owners or advisers in your area, I speak on this at [business keynotes](https://theaieducator.io/business-keynotes?utm_source=business.theaieducator.io&utm_medium=referral&utm_campaign=business-insurance-cover-ai-mistakes).

## Sources and further reading

- Trades Coverage, [The AI exclusion hiding in your contractor insurance](https://tradescoverage.com/reports/genai-exclusion-adoption/), August 2026 (data through 31 July 2026; state insurance filings in SERFF and Florida I-File).
- Verisk via Independent Insurance Agents & Brokers of America, [Verisk to Roll Out New General Liability Exclusions for Generative AI Exposures](https://www.independentagent.com/vu_resource/verisk-to-roll-out-new-general-liability-exclusions-for-generative-ai-exposures/), forms with a January 2026 edition date.
- Gallagher, [ISO Introduces Generative AI Exclusion in Commercial General Liability Policies](https://www.ajg.com/news-and-insights/iso-introduces-generative-ai-exclusion-in-commercial-general-liability-policies/), July 2026.
- Insurance Business, [AI exclusions are spreading through commercial policies, brokers warn clients to check cover](https://www.insurancebusinessmag.com/us/news/professional-liability/ai-exclusions-are-spreading-through-commercial-policies-brokers-warn-clients-to-check-cover-594489.aspx), 2026 (date not shown on the page read).
- Resultsense, [London insurer covers AI risk as US carriers exclude it](https://www.resultsense.com/news/2026-07-30-cfc-affirmative-ai-cover-lloyds-market/), 30 July 2026.
- DAC Beachcroft, [AI and Policy Wordings](https://lmalloyds.com/wp-content/uploads/2026/04/AI-and-Policy-Wordings-DACB-21-April-2026.pdf), presentation to the Lloyd's Market Association wordings forum, 21 April 2026.

*Dan Fitzpatrick is a keynote speaker, Forbes contributor and five-times bestselling author who has trained more than 150,000 educators across 30+ countries. [About Dan](https://theaieducator.io/about).*

## Frequently asked questions

### Does business insurance cover mistakes made by AI?

Sometimes, and it depends on your wording. Since January 2026, US insurers can attach standard endorsements that exclude losses arising from generative AI, and some London insurers now write affirmative AI cover. Only your policy documents settle it, so ask your broker to confirm in writing which forms apply, with form numbers.

### What are CG 40 47, CG 40 48 and CG 35 08?

They are three optional ISO generative AI exclusions for commercial general liability policies, edition January 2026. CG 40 47 covers bodily injury, property damage and personal and advertising injury. CG 40 48 covers personal and advertising injury only. CG 35 08 covers products and completed operations. An insurer must choose to attach them.

### How many insurers have filed AI exclusions?

Trades Coverage found 4,078 filed form adoptions across 49 US states and the District of Columbia, with data through 31 July 2026, up from zero in mid-2025. Filings show what carriers may use, not what sits on your policy, so the count signals direction and is not a measure of your own cover.

### Is UK business insurance treating AI differently?

For now, yes in places. Resultsense reported in July 2026 that CFC added affirmative AI wording across seven products, while the Lloyd's Market Association was waiting for a market instruction before drafting exclusions. I found no UK survey counting exclusions on small business policies, so check your own wording.

### Which policy responds when AI makes a mistake?

It depends on what went wrong. A broker quoted by Insurance Business said general liability, errors and omissions, cyber and management liability could all be triggered, depending on the circumstances. A wrong estimate, a chatbot promise and an AI-drafted contract can each land on a different policy, so map each use.

### What should a small business ask its broker about AI?

Ask three things: does any policy carry a generative AI exclusion, does any carry affirmative AI wording, and which policy would respond to each way you use AI. Send your AI Use Map, ask for form numbers, and get the answer in writing. A verbal reassurance is not something you can rely on at claim time.

---
Source: [Does Business Insurance Cover Mistakes Made by AI?](https://business.theaieducator.io/posts/business-insurance-cover-ai-mistakes)
Publisher: [The AI Educator](https://theaieducator.io)
