AI Insurance
If your business relies on AI and its output is wrong, who pays? The liability sits with you, not the AI vendor and not the machine. AI insurance ensures your policies respond when that happens.
Quick Summary
- The liability is yours: When AI produces a flawed report, a defective design or an incorrect decision, the claim lands on the business that used it. AI vendors’ terms of use almost universally exclude liability for the consequences of their output.
- Most policies are silent on AI: The majority of professional indemnity, cyber and liability policies neither confirm nor exclude AI-related claims. That uncertainty is only resolved at the point of a claim, which is the worst possible time.
- Affirmative cover removes the doubt: A small number of insurers now state explicitly, in the policy wording, that claims involving AI are covered. Daines Kapp places business directly with CFC and Hiscox, both of which have introduced affirmative AI wording.
- Whole-programme review: AI exposure does not sit neatly in one policy. It crosses professional indemnity, cyber, directors’ and officers’, media and product liability, so we review the programme as a whole.
AI Is Now in Your Business, Whether You Bought It or Not
Most UK businesses now use AI in some form: drafting documents and advice, analysing data, generating designs and marketing content, answering customer enquiries, or embedded invisibly inside software the business already licenses. Very few have asked whether their insurance would respond if that AI produced something wrong and a client suffered a loss as a result.
At Daines Kapp we have been advising on technology-driven risk for years. AI and cyber risk is led in-house by Stefan Daines, a recognised speaker on AI and insurance risks. We bring the discipline of a BIBA Accredited Cyber Insurance Broker to AI: understand the exposure first, then place cover that states plainly how it will respond.
Who Is Liable When AI Gets It Wrong?
In almost every case: the business that used it. If a consultancy relies on an AI tool to produce analysis for a client and the analysis is flawed, the client claims against the consultancy. If a manufacturer uses AI in design or quality control and a defective product reaches the market, the product liability claim comes to the manufacturer. The AI vendor sits behind terms of use that exclude responsibility for how the output is used, in the same way that cloud and IT providers exclude liability for data breaches.
This mirrors a lesson our clients already know from cyber: outsourcing a function does not outsource the legal liability. Using a third party’s AI does not transfer your duty of care to your own clients.
Silent Cover vs Affirmative Cover: The Question That Matters
One of the most crucial distinctions between insurance policies is between those which specify AI coverage and those which do not.
Silent cover means the policy wording says nothing about AI either way. A claim involving AI might be paid under the policy’s general terms, or it might not. The answer depends on interpretation, argued out after the loss has happened.
Affirmative cover means the policy states explicitly that claims involving AI are covered: for example, where a business has relied on an AI tool’s output to deliver a contract and the output was wrong, or where advice on an AI system proves negligent.
In practice the market now holds three positions rather than two. Some insurers exclude AI-related claims outright. Some provide affirmative cover, either as standard or by endorsement. Many remain silent, which is still the most common position in the UK.
The move towards exclusion is clearest in the United States, where a number of carriers have begun narrowing their wordings specifically to avoid repeating the silent cyber problem of the last decade. It would not be surprising to see that filter into the UK in time, not least because several providers writing in the US also supply capacity into the UK market.
The movement in the other direction is real too. Hiscox introduced affirmative AI wording into its Technology Professional Indemnity policy in 2025, reported at the time as the first in the UK market to grant clear cover for AI-related claims. CFC has embedded affirmative AI wording across its product range, including professional liability, management liability, media, technology and cyber, addressing exposures such as AI-generated content and model errors.
Two policies at a similar premium may respond entirely differently to the same AI-related claim. Knowing which of those three positions your programme sits in, before a claim, is what a specialist broker is for.
Why AI Cover Is Hard to Price, and Why That Matters to You
Insurers price most risks from history. They know roughly how often warehouses catch fire and what it costs when they do. AI offers no such record. The claims history is short, the systems change between one renewal and the next, and there is no way to inspect from the outside how a business actually uses AI, in the way a surveyor can inspect a building.
The consequence is that insurers are currently pricing uncertainty rather than measured risk, and they are managing that in different ways. Some by excluding. Some by asking a great deal more at the proposal stage. Some by writing affirmative cover and accepting what they cannot yet measure.
Three things follow for you as a buyer. Two quotes at a similar price may reflect completely different appetites, so this year the wording deserves more attention than the premium. The questions at renewal are getting longer and more technical, which makes a vague answer a pricing problem rather than an administrative one. And a business that can describe its own AI use clearly is in a materially stronger position than one that cannot, which is preparation rather than luck.
Where AI Risk Sits in Your Insurance Programme
AI is not a single risk with a single policy. The table below maps the main exposures to the policies that should address them.
| Policy Line | AI Exposures It Should Address | Daines Kapp Advantage |
|---|---|---|
| Professional Indemnity | Claims arising from advice, reports, designs or analysis produced with AI assistance, including errors caused by relying on flawed AI output | We check whether your PI wording is silent or affirmative on AI, and can place cover with insurers whose wordings state explicitly that AI-assisted work is covered |
| Cyber | AI-enabled attacks such as deepfake fraud and AI-crafted phishing; security failures in AI tools holding your data; breaches of data fed into AI systems | As a BIBA Accredited Cyber Insurance Broker we assess whether your cyber policy’s crime and social engineering sections respond to AI-enabled fraud |
| Directors’ & Officers’ | Claims against directors for decisions to adopt (or ignore) AI, governance failures around AI use, and regulatory investigations into how AI was deployed | AI governance is becoming a board-level duty; we ensure management liability cover reflects it |
| Media Liability | Defamation, copyright and intellectual property claims arising from AI-generated content in marketing, publishing and communications | We place with insurers whose media wordings confirm that AI involvement in content does not prevent the policy responding |
| Product Liability | Injury or damage caused by products designed, tested or controlled using AI, and by products with embedded AI functions | We ensure the policy accurately describes the role AI plays in your production and design processes, so nothing is inadvertently excluded |
Who We Help
Our starting point is not “AI companies”. It is ordinary businesses whose work now depends on AI, often more than they realise:
- Professional services firms: consultancies, surveyors, engineers, accountants and other advisers using AI to produce or check client work. The output carries your name and your liability.
- Manufacturers: businesses using AI in design, quality control, predictive maintenance or embedded product functions, where an AI failure becomes a product liability or recall issue.
- Construction and property businesses: firms adopting AI in surveying, design, estimating and project management, where errors flow into contracts and the built environment.
- Media, marketing and creative businesses: agencies and publishers using generative AI, where content risk (copyright, defamation, intellectual property) has changed shape.
- Any business with customer-facing AI: chatbots and automated decision tools that give information or make commitments on your behalf.
Common Misconceptions About AI and Insurance
“The AI provider is liable, not us.” Read their terms of use. AI vendors exclude liability for the accuracy of output and its consequences, just as IT contracts commonly cap the provider’s liability, sometimes at a fixed sum and sometimes at a multiple of the fees paid. Your clients will claim against you, and your contract with the vendor will not bridge the gap. Insurance is what bridges it.
“Our existing PI will cover it.” Perhaps. Most PI policies were written before generative AI existed and are silent on it. Silent cover means the answer is decided by interpretation after the loss. Affirmative cover means the answer is written into the policy before it. If your work now involves AI, that difference is worth resolving at renewal, not at claim.
“We only use AI for admin, so there’s no exposure.” Businesses consistently underestimate where AI is in use. Staff adopt tools without approval, and AI is embedded in mainstream software by default. If any AI-touched output reaches a client, a product, or a published channel, the exposure exists.
Why Choose Daines Kapp for AI Insurance?
Daines Kapp is a BIBA Accredited Cyber Insurance Broker, and our cyber and AI advice is led by Stefan Daines, a recognised speaker on AI and insurance risks. We were among the first UK brokers to publish analysis on what frontier AI development means for UK businesses and their insurance.
We place business directly with CFC and Hiscox, both of which have introduced affirmative AI wording, and with Cowbell, which has built AI-specific factors into the way it underwrites cyber. We also have access to the wider specialist market as a Willis Network Broker.
AI risk does not respect policy boundaries, so we do not review it one policy at a time. We look at your programme as a whole: what AI is actually in use across the business, which policies are silent, which are affirmative, and where the gaps sit. Then we present the position in plain English so you can make an informed decision.
Frequently Asked Questions About AI Insurance
What is affirmative AI cover?
Affirmative AI cover means the insurance policy states explicitly, in its wording, that claims involving artificial intelligence are covered. This contrasts with “silent” cover, where the policy neither mentions nor excludes AI, leaving the response to be argued at the point of a claim. Affirmative wordings typically confirm cover for losses where a business relied on AI output to deliver its work, where AI it built or advised on failed to perform, or where AI contributed to content that gave rise to a claim. In the UK, Hiscox introduced an affirmative AI clause in its technology professional indemnity policy, and CFC has rolled affirmative AI wording across its wider product range.
Does professional indemnity insurance cover mistakes made using AI?
It depends entirely on the wording. Most professional indemnity policies do not mention AI, so a claim arising from AI-assisted work falls to be considered under the policy’s general terms. In many cases it may be covered, because the claim is ultimately about negligent professional work regardless of the tool used, but that is not guaranteed, and some insurers are beginning to introduce AI exclusions. If AI now plays a material role in producing your client work, the safest position is a wording that addresses AI affirmatively. We review your existing PI wording and advise whether it needs to change.
Who is legally responsible if AI gives wrong information to my customers?
Your business is. Whether the wrong information comes from an employee, a spreadsheet or an AI chatbot, the duty to your customer is yours and the claim will be made against you. AI vendors’ contracts exclude liability for the accuracy of output, so recovering the loss from the tool’s provider is rarely realistic. This is why AI exposure needs to be addressed within your own insurance programme rather than assumed to sit with the technology supplier.
Do I need a separate AI insurance policy?
Usually not. For most businesses, AI risk is best addressed by ensuring the existing programme (professional indemnity, cyber, directors’ and officers’, media and product liability) responds properly to AI-related claims, ideally through affirmative wordings. Standalone AI policies do exist, but it is worth knowing who they are built for. Most are aimed at the developers and vendors of AI systems rather than the businesses using them, and they typically take the form of performance warranties that respond when a model fails to perform as promised. Most of that activity is currently in the United States. If your business uses AI in its work rather than sells it, fixing the existing programme is almost always the better answer. The starting point either way is a review of what AI is in use and how your current policies would respond.
What will insurers ask about our use of AI?
Expect questions about which AI tools the business uses and for what, whether AI output is reviewed by a person before it reaches clients or customers, what data is put into AI systems, and whether the business has an AI use policy for staff. Insurers are not looking for perfection; they are looking for awareness and governance. This is becoming formal rather than conversational. Cowbell introduced dedicated AI underwriting factors in 2026 that assess visibility of AI use across a business, how autonomously its AI systems operate, and the governance around them. A business that can describe its AI use clearly presents far better to underwriters than one that cannot. We help clients prepare this picture before approaching the market, which supports both terms and pricing.
