llm-integrated

Definition

Describes an application that calls a language model inside its own workflow — a mail client, a search box, an internal tool — as opposed to a standalone chatbot. The phrase is common in security papers because integration is precisely where the attack surface appears: the application feeds the model third-party content, and that content can be read as instructions. The risk comes from the plumbing, not the model.

Explainers using this term

• LLM Security — Prompt Injection and How to Actually Defend Against ItNot What You've Signed Up For: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection