Prompt injection — Glossary

Prompt injection exploits the fact that instructions and data share the same natural-language channel. Malicious content in a user message or a retrieved document can try to override system rules—“ignore previous instructions…”—or trigger privileged tools.

Defenses are architectural: least-privilege tools, server-side authorization, human approvals for side effects, and treating retrieved content as untrusted. A single system-prompt warning is not enough.

See the encyclopedia topic Prompt Injection.