Prompt Injection
Prompt injection is an attempt to redirect an AI system through instructions embedded in input or external material. The system may treat that material as authoritative instructions instead of data to process.
[OWASP]In practice · hypothetical example
A retrieved page contains a demand to ignore the user’s task. The application should treat that text as untrusted page content.
[OWASP]A little deeper
Indirect injection can arrive through retrieved documents or tool results. Controls must consider how untrusted content reaches the model and what actions the surrounding application permits. [OWASP]
A common mix-up
Only the user’s direct message can contain an injection.
External content can also carry an instruction attempt. [OWASP]
Can a retrieved document carry prompt injection?
Sources & editorial notes
Evidence: supported. Primary-source support for this scoped entry; publication approved by the project owner.
- LLM01:2025 Prompt Injection ↗ (opens in new tab)OWASP · Publication date unknown
Relevant section: Types of prompt injection; Prevention and mitigation strategies
Last editorial review: 2026-09-13 by project-owner.
First observed in this corpus: Unknown.
Revision history
Revision 2 · Created 2026-09-13 · Updated 2026-09-13
Project owner approved the current content for publication. Existing evidence scope and limitations remain applicable.