AI exclusions and clauses in insurance policies, explained
What is an AI exclusion?
An AI exclusion is a clause that removes cover for claims arising from artificial intelligence — typically claims “arising out of, based upon or attributable to” the use, development or deployment of AI systems. The broadest versions, sometimes called absolute AI exclusions, sweep in any claim with an AI connection anywhere in its causal chain.
The breadth is the problem. AI is now embedded in everyday software — email filters, drafting tools, analytics, design packages — so an aggressively worded exclusion could in principle reach claims that have only an incidental AI element. If an AI exclusion appears in your renewal papers, the questions to ask are: how does it define AI, does it require AI to be the proximate cause or merely present, and does it distinguish between your deliberate use of AI and AI buried in tools you use incidentally. The answers vary enormously between wordings, which is precisely why the clause has to be read rather than assumed.
What is affirmative AI cover?
Affirmative AI cover is the opposite response: wording that states expressly that claims arising from AI use are covered, sometimes with its own definitions, conditions and sub-limits. It appears both as endorsements to conventional PI and cyber policies and, increasingly, as standalone AI liability products aimed at firms that build or deploy AI at scale.
Express cover is generally good news — certainty in both directions beats ambiguity. But the details carry the value: affirmative AI cover may sit behind a lower sub-limit than the main policy, may be conditional on stated governance controls such as human oversight, and may define AI narrowly enough to leave some tools outside. Affirmative cover is worth having; it is also worth reading.
What is silent AI?
Silent AI describes the position most policies are actually in: the wording neither mentions nor excludes AI, because it was drafted before generative AI existed. Claims involving AI then fall to be decided under general policy language — definitions of professional services, standard exclusions, conditions about supervision and disclosure.
Silence can work for the policyholder: a negligent report is a negligent report however it was drafted, and general wording will often respond. But silence is ambiguity, and ambiguity is what coverage disputes are made of. Insurers dislike it too — it leaves them exposed to risks they never priced — which is why the market is moving, wording by wording, from silence towards express positions in both directions. Every renewal is a chance for your policy’s position to change without ceremony. Renewal documents need reading for exactly this reason.
Where these clauses are appearing
Professional indemnity is the front line, because AI is changing how professional work is produced. Cyber policies are addressing AI from the threat side — AI-enabled attacks — and some now speak to losses involving deepfake-enabled fraud, in either direction. Directors’ and officers’ policies meet AI through management decisions: statements to investors about AI capability, board oversight of AI risk. Media and IP-flavoured wordings meet it through AI-generated content. The common thread is that each class is deciding separately how to respond, so a firm’s programme can hold different AI positions in different policies at the same time. Checking them together, as a programme, is the point of a broker review.
What to do about it
Three steps, none of them dramatic. Read the current wordings — all of them, since positions differ policy by policy — specifically for AI language, and diff each renewal against the expiring wording, because exclusions arrive quietly. Disclose your AI use accurately when asked, and materially significant use even when not asked. And where your work genuinely depends on AI, ask the question directly through your broker: will this policy respond to an AI-related claim, and if the answer is unclear, find a wording where it is not.
Frequently asked questions
Do all insurance policies now exclude AI?
No. Express AI exclusions exist and are becoming more common, but most policies currently in force say nothing about AI either way, and a smaller group expressly cover it. The three positions — excluded, covered, silent — coexist in the market, sometimes within one firm’s own insurance programme.
Is silent AI cover better than an AI exclusion?
Usually, but it is not safety — it is ambiguity. Under a silent wording a claim involving AI is argued under general policy language, and the outcome depends on the facts and the definitions. An express position, ideally affirmative cover with sensible conditions, beats both silence and exclusion.
Can an AI exclusion be added at renewal without us noticing?
Yes, in practice. Renewal is a new contract, and wordings change year to year. New exclusions appear in the renewal documentation, not in a separate warning letter. Comparing the renewal wording against the expiring one — something a broker should do as a matter of course — is how they get caught.
What is an absolute AI exclusion?
The broadest form of AI exclusion: one that removes cover for any claim arising out of or attributable to AI, without requiring AI to be the main cause. Because AI is embedded in ordinary business software, wordings this wide can reach claims with only an incidental AI connection — which is why the definition and causation language deserve close reading.
Apex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority (FRN 724952). This page is general information, not advice on a specific policy.