Learn

Learn the core ideas behind prompt injection, hidden instructions in documents, findings and coverage, and how to review content before using it with an AI system.

Start here

These guides explain what the scanner is designed to help you review. They describe public product concepts without exposing private detector rules or test material.

What is prompt injection?

Understand how untrusted content can conflict with the task you intended an AI system to perform.

Hidden instructions in files

Learn why documents can contain relevant content outside the visible body text and why file structure matters.

How to review findings

Learn how to interpret findings, coverage and limitations before deciding what to do with scanned content.

Use Learn with the scanner

Learn provides context for the scanner, not a replacement for reviewing the original content. Use findings together with coverage, compare them with the source document, and consider the permissions of the AI system that may receive the content.

Glossary