Glossary.

Acceptable-Use PolicyAUP

A short rulebook that says what staff are and are not allowed to do with AI tools at work. Think of it as the "house rules" for AI. There's a ready-made one in the templates.

AI tool

A program you give words to, and it writes, answers, or makes something back. ChatGPT, Copilot, Claude and Gemini are AI tools. They sound clever, but they don't truly "know" things, they predict what to say next.

Confabulationaka hallucination

When an AI makes something up but says it with total confidence, a fake fact, a made-up quote, a court case that never happened. The AI isn't lying on purpose; it just fills gaps with words that look right. Always check important answers.

Consumer AI

The free or personal version of an AI tool, like the one you'd use at home. It's handy, but the company often gets to keep and learn from what you type. Not the place for work secrets.

Data breach

When private information gets out to people who shouldn't see it, by accident or by attack. Pasting a customer list into the wrong AI tool can count as one. Breaches can mean fines and lost trust.

Data controller

The business that decides why and how people's personal information is used. Under the law, the controller is responsible for keeping it safe, even when a tool does the actual work.

Data processor

A company that handles personal information on behalf of the controller, for example, an AI vendor. They must follow the controller's instructions and keep the data safe.

Deepfake

A fake video or voice made by AI that looks or sounds like a real person. Scammers use them to pretend to be your boss or a customer. If a "video call" asks you to send money, stop and check another way.

Enterprise AI

The paid, business version of an AI tool. It usually promises not to train on your data and keeps things more private. Safer for work, but it still isn't a free pass to paste anything.

EU AI Act

A European law about using AI responsibly. One part (Article 4) says staff who use AI need to understand it well enough to use it safely, which is exactly what this course gives you.

Human oversight

Keeping a real person in charge of what AI does, checking its work and making the final call. The AI helps; the human stays responsible.

ISO/IEC 42001

A newer worldwide standard, this one for managing AI responsibly. It helps a business show it uses AI in a safe, organised way.

Personal dataPII

Any information that points to a real person, name, email, phone number, address, health or pay details. This is the stuff you must protect most carefully, and rarely paste into AI.

Phishing

A trick message that pretends to be someone you trust, to make you click, pay, or hand over a password. AI now makes these messages neat and typo-free, so judge the request, not the spelling.

Prompt

The words you type into an AI tool to tell it what you want. "Write a polite reply to this email" is a prompt.

Prompt injection

A sneaky attack where hidden instructions in a webpage or document trick the AI into doing something it shouldn't, like leaking information. A reason not to fully trust AI with secret data.

Retention

How long a company keeps what you typed into its AI tool. "Deleting the chat" doesn't always mean it's really gone, check the tool's settings.

Shadow AI

Using AI tools at work that nobody approved or knows about, a personal account, a random browser add-on, an app with AI hidden inside. It feels harmless but it's where most accidents start.

Training data

The huge pile of text an AI learned from. If a tool "trains on" what you type, your words could become part of how it answers other people, so don't feed it secrets.

Two-source rule

Before you trust an important AI answer, check it against a second, reliable source. If only the AI says it, treat it as "maybe", not "true".