Why it matters for your company
Limited exclusions are narrower by design. They name a type of system, a type of output or a coverage part, and they often carve back the insured's internal use of AI tools. The standard general liability versions released for January 2026 are limited in this sense (Independent Agent magazine, 2025-10-21).
Narrow wording can still reach far. A definition written around content generation may be read to include an agent's actions or a robot's perception model, and commentators have flagged how uncertain that reach is (Pillsbury, 2026-04-13). Reading the definition against what the product does is the only way to judge it.
Related terms
- Absolute AI exclusionAn exclusion that removes any claim arising from, relating to or involving artificial intelligence in any way, with no carve-back for the insured's ordinary use. For a company whose product is a model or a machine, it can leave the policy close to empty.
- Generative AI exclusionAn endorsement that removes coverage for claims arising out of generative artificial intelligence. Standard-form versions for general liability, including one for products and completed operations, took effect in January 2026.
- Silent AIAI-related risk that a policy neither expressly covers nor expressly excludes, so whether it responds depends on how general wording is read after a loss. The term borrows from silent cyber.
- EndorsementA page attached to a policy that changes it: it adds coverage, removes it, names a new party, or rewrites a definition. The endorsement controls over the base form wherever the two disagree.