# Prompt injection

Text that makes an AI model do something its owner did not intend.

OWASP adds that the text need not be readable by a person: injected inputs "do not need to be human-visible/readable, as long as the content is parsed by the model". An agent that reads outside content or has tools is more exposed, because injected text can reach its [tool calls](/reference/glossary/tool-call).

What guardrails catch and miss: [Can you stop an AI agent being tricked?](/research/ai-agent-security-and-guardrails)

---

Published 2026-11-06. Web page: https://www.mackenziebowes.com/reference/glossary/prompt-injection
