AI insurance glossary
Plain-language definitions of the terms used across this guide. Informational only, not insurance or legal advice.
Insurance runs on precise words, and AI coverage adds a second layer of jargon on top. These definitions are written for buyers who want to read a policy or a carrier's product page without a dictionary open beside them. Where a term shapes a real coverage decision, the entry links to the tool or article that goes deeper.
Coverage grants and gaps
- Affirmative AI coverage
- Policy language that names AI risk and states clearly that it is covered, rather than leaving the question open. Affirmative language is what a buyer wants to see in writing.
- Silent (non-affirmative) coverage
- A policy that neither grants nor excludes AI risk explicitly. Whether a loss is covered gets decided at claim time, often in dispute, which makes silence the riskier position.
- AI exclusion
- Language that removes coverage for losses arising from the use, development, or output of artificial intelligence. Scope varies widely; the exclusion tracker follows how carriers word these.
- Carve-back
- Language that restores a slice of coverage inside a broader exclusion, giving back a defined set of otherwise-excluded losses. Reading the carve-back is as important as reading the exclusion.
Policy types and structures
- Errors and omissions (E&O)
- Professional liability insurance that responds to claims of negligent work, mistakes, or failure to deliver a professional service. Whether it reaches AI losses is the question behind our E&O gap page.
- Technology E&O
- A form of E&O built for technology companies, covering financial harm to clients from failures of a product or service.
- Standalone AI policy
- A policy purpose-built for AI risk rather than an add-on to a general or professional program. It may offer broader or clearer terms; compare types on the carrier comparison.
- Endorsement
- An amendment attached to a policy that adds, limits, or clarifies coverage, including endorsements that specifically address AI.
- Difference-in-conditions (DIC) wrap
- A policy layered over primary coverage to fill gaps the underlying policies leave open.
- Performance guarantee
- A promise, often from a vendor, that a product will meet a stated standard, with the vendor making the buyer whole if it does not. It is not the same as insurance.
- Indemnity
- A regulated promise from a licensed insurer to pay covered losses under a policy. The difference between a guarantee and an indemnity is who stands behind the money.
Mechanics of a policy
- Claims-made policy
- Coverage that responds to claims first made during the policy period, subject to any retroactive date.
- Occurrence policy
- Coverage that responds to events that happen during the policy period, whenever the claim is later reported.
- Retroactive date
- The earliest date of wrongful acts a claims-made policy will cover. Losses tied to earlier acts fall outside the policy.
- Duty to defend
- An insurer's obligation to provide and pay for the defense of a covered claim, sometimes in addition to the limits.
- Aggregate limit
- The most a policy will pay for all covered losses during the policy period.
- Retention (deductible)
- The amount the insured pays on a claim before the insurer's coverage begins.
- Underwriting
- The process by which an insurer evaluates a risk and decides whether, and on what terms, to offer coverage. See the coverage readiness tool for the questions underwriters ask.
AI-specific risk terms
- Model drift
- The gradual decline in an AI model's accuracy as real-world data moves away from the data it was trained on. Drift can turn a once-reliable system into a source of loss.
- Hallucination
- Confident but false or fabricated output from a generative AI system, a common source of AI liability concern.
Have a question this glossary does not answer? The FAQ covers the questions buyers ask most, and the terms and disclaimer explain how to use this guide.