ITG NowHome › AI glossary › Prompt injection
What is Prompt injection?
Prompt injection is an attack where hidden text tricks an AI into following the attacker's instructions instead of yours.
- Category Safety & law
- Updated 20 Oct 2026
Prompt injection, explained simply
AI reads text and treats much of it as instructions. Prompt injection hides instructions in a web page, email or file.
When the AI reads that content, it may follow the hidden text. That could make it leak information or take a wrong action.
This matters most for AI agents that browse the web or read your mail. Treat what an agent reads as untrusted.
On ITG Now, our free tool guides and daily news stories flag when an agent tool reads outside content and what that risk means.
Related terms
JailbreakAI agentGuardrailsBrowser agentTool callingAI safety
Questions people ask about Prompt injection
Prompt injection vs jailbreak: what is the difference?
A jailbreak comes from the user who wants the model to break its rules. Prompt injection comes from hidden text in content the AI reads.
How can I stay safe from prompt injection?
Give agents only the access they need and ask them to confirm important actions. Be careful with unknown pages and files.
Can prompt injection be fully stopped?
Not yet. Makers add defenses, but no method blocks every case, so caution still matters.