AI insurance glossary

44 terms you will meet in AI insurance wording, quotes and broker emails, each in a sentence or two, with links to the pages that go deeper.

Cover and exclusions

Affirmative AI cover: Policy wording that says in so many words that claims involving AI are covered, instead of leaving the policy silent. It is often added by endorsement, sometimes with its own sublimit. See AI liability insurance.
Silent AI: A policy that neither covers nor excludes AI by name. An AI-caused claim is then handled like any other claim under the policy's terms, which can leave room for disputes. The term borrows from 'silent cyber'.
AI exclusion: A clause or endorsement that removes claims arising out of AI from a policy. How far it reaches depends on how the policy defines AI. The AI exclusions tracker logs the forms and insurers using them.
Absolute AI exclusion: An exclusion that removes any claim connected to the use, deployment or development of AI, with no carve-backs. A May 2025 law-firm alert described one used by W. R. Berkley in D&O, E&O and fiduciary policies.
ISO generative AI exclusions (CG 40 47, CG 40 48, CG 35 08): Optional general liability endorsements from Verisk's ISO unit, edition-dated January 2026, that insurers can attach to remove claims arising out of generative AI. Paste a forms schedule into the AI clause scanner to spot them.
Endorsement: A document attached to a policy that adds to, removes or changes its cover. Both AI cover and AI exclusions are commonly added this way.
Carve-back: Wording that restores part of the cover an exclusion took away, often starting 'notwithstanding' or 'however, this exclusion does not apply to'.

Policy mechanics

Sublimit: A cap inside the overall policy limit that applies to one kind of loss. For example, Embroker caps algorithm-removal costs under its AI endorsement at $150,000.
Retention (deductible): The part of a covered loss the policyholder pays before the insurer pays. Higher retentions usually mean lower premiums.
Claims-made policy: A policy that responds to claims first made during the policy period, for acts after its retroactive date. Most tech E&O, cyber and D&O policies are written this way.
Occurrence policy: A policy that responds to injury or damage that happens during the policy period, whenever the claim is later made. Most general liability policies work this way.
Retroactive date: The earliest date of a wrongful act that a claims-made policy will cover. Work done before it is outside the policy even if the claim arrives later.
Duty to defend: An insurer's obligation to defend a covered claim, usually by appointing and paying lawyers. Some policies instead reimburse defense costs the policyholder has paid.
Difference in conditions (DIC): Cover that sits over or alongside existing policies to fill gaps in their terms. The Mayflower Specialty and Hadron program includes an AI difference-in-conditions excess layer.
Excess layer: Insurance that pays once a primary policy's limit is used up, adding more limit on top.
First-party and third-party cover: First-party cover pays your own losses, such as the cost of taking a faulty AI system offline. Third-party cover pays claims others bring against you, such as a client suing over an AI error.

Lines of insurance

Tech E&O (technology errors and omissions): Pays claims from clients who say your technology product or service failed them and cost them money, including defense costs. Often sold in one policy with cyber. See does tech E&O cover AI.
Professional liability: The broader name for errors and omissions cover for professional services. Whether work delivered with AI counts as a 'professional service' depends on the policy's definition.
Cyber insurance: Covers losses from data breaches, network security failures and cyber crime: your own response costs and claims brought against you. Some cyber policies now name AI security events and deepfake-instructed payments.
D&O (directors and officers) insurance: Covers claims against directors and officers over how they ran the company and, in many policies, securities claims against the company itself. Overstated AI claims to investors, known as AI washing, are a D&O exposure.
EPL (employment practices liability): Covers claims from employees or job applicants over discrimination, harassment or wrongful termination, including claims about biased AI hiring tools.
Media liability: Covers claims over content you publish, such as defamation, copyright or privacy claims. AI-generated content raises the same questions.
Social engineering fraud cover: Crime or cyber cover that pays when an employee is tricked into sending money to a fraudster. It is often sublimited, and some policies now name deepfake impersonation. See deepfake fraud insurance.
Funds transfer fraud: Cover for a fraudulent instruction sent to your bank to move money without your knowledge. It differs from social engineering cover, where your own employee is deceived into making the payment.
AI performance guarantee: Cover or a warranty that pays when an AI model fails to meet agreed performance thresholds, as with Munich Re's aiSure and Armilla Guaranteed.

AI risks

Hallucination: When a generative AI system produces false or invented output and presents it as fact. See AI hallucination insurance.
AI agent: An AI system that takes actions on its own, such as sending messages, changing records or making payments through connected tools. See AI agent insurance.
Algorithmic bias claim: A claim that an automated system treated people unfairly on a protected characteristic, for example in hiring, lending or insurance pricing.
Deepfake: AI-generated or AI-altered audio, video or images that impersonate a real person. Used in payment fraud and fake job candidates. See deepfake fraud insurance.
AI washing: Overstating or misrepresenting what a product's AI does, to investors or customers. US regulators have brought enforcement actions over it. See the AI washing tracker.
Model drift: A fall in a model's accuracy over time as the data it sees moves away from the data it was trained on.
Vendor IP indemnity: A promise by an AI provider to defend customers against intellectual property claims over AI output, usually only on paid tiers and only if conditions are met. See the AI vendor indemnity tracker.

Standards and rules

AIUC-1: A certification standard for AI agents published by the Artificial Intelligence Underwriting Company (AIUC), designed to be paired with insurance. See AIUC-1 explained.
ISO/IEC 42001: The international standard for an AI management system, published in December 2023. Certification is one way a company can show underwriters how it governs AI.
NIST AI RMF: The US National Institute of Standards and Technology's voluntary AI Risk Management Framework, released in January 2023.
EU AI Act: The European Union's AI law, Regulation (EU) 2024/1689. It entered into force on 1 August 2024 and applies in stages, with duties that rise with an AI system's risk level.

The market

Underwriting: How an insurer assesses a risk and sets terms and price. For AI cover this can include reviewing the AI system itself as well as the business.
Supplemental application: An extra questionnaire about one risk, such as AI use, filled in alongside the main insurance application.
MGA (managing general agent): An intermediary with authority from an insurer to underwrite and issue policies on that insurer's paper. Many new AI insurance products are written this way.
Capacity: The amount of insurance an insurer, reinsurer or Lloyd's syndicate is willing to write. MGAs and startup insurers rely on capacity from others.
Admitted insurer: An insurer licensed by a state's insurance regulator to sell in that state, with its forms and rates subject to state rules.
Surplus lines (E&S): Insurance from non-admitted insurers for risks the standard market will not take, placed through a licensed surplus lines broker. New and unusual covers, including several AI products, are often sold this way.
Broker: A licensed intermediary who places insurance for the buyer, usually paid by commission from the insurer. An agent, by contrast, represents one or more insurers.
Lloyd's syndicate: A group of members that underwrites risks at Lloyd's of London. Several specialist AI products are backed by Lloyd's capacity.

Missing a term? Email it to us and we will add it. To check your own policy for AI wording, use the AI clause scanner.