AI glossary

Prompt injection

Also called indirect prompt injection.

Prompt injection is an attack that plants hidden instructions for an AI inside something it reads, such as an email, a document or a web page, so that the AI follows the attacker instead of you. The UK’s National Cyber Security Centre says it may never be fully fixed, because today’s models don’t reliably separate instructions from data.

Think of it as

A note in the in-tray saying “ignore your manager”.

Why it matters at work

It matters most once AI can act through connectors or agents, because then hidden words can send, delete or share things. Limit what the tool can do rather than relying on it to spot the trick; Module 8 shows how.

Sources

Prompt injection is not SQL injection (it may be worse) (opens in a new tab) National Cyber Security Centre, 8 December 2025.