AI Safety Basics — AI Basics

You'll learn

The three starter habits that prevent most AI mistakes
What's fine to share with an AI tool — and what never is
A quick routine for double-checking outputs before they leave your hands
The human-stays-responsible rule

01Section

Safety is three habits, not a rulebook

"AI safety" sounds like a heavy topic — and parts of it are. But for everyday use, it comes down to three habits: share carefully, check before you rely, and own what goes out under your name.

This lesson is the starter version. Each habit takes seconds once it's automatic, and together they cover the mistakes beginners actually make. When AI starts touching sensitive work — client data, regulated fields, publishing at scale — there's a whole path that goes deeper. We'll point you there at the end.

Not about fear

These habits are the seatbelt, not a reason to stay off the road. You buckle up in seconds and then drive normally.

Checkpoint

02Section

Habit one: share carefully

Typing into an AI chat is typing into someone else's website. Most of what you'd tell it is fine: general questions, descriptions of situations, drafts meant for other people's eyes anyway. Some things are never fine:

Passwords, login codes, and account numbers — no exceptions, no matter how helpful the task seems.
Other people's private information: a customer list, a coworker's situation with their name attached, anyone's medical or financial details.
Confidential work material — contracts, internal documents, anything under NDA — unless your employer has approved the tool for it.
Anything you'd be uncomfortable seeing outside the chat.

Privacy note: The rule of thumb: if you wouldn't pin it to a public noticeboard with your name on it, don't paste it. You can almost always describe the situation without the identifying details — and the AI's answer will be just as good.

Describing instead of pasting

Instead of pasting an unhappy client's actual email — name, company, order number — write: "A long-time client is upset about a two-week delay on their order. I want to apologize honestly, explain the cause briefly, and offer a concrete next step. Draft a short, warm reply." Same quality of help, nothing exposed.

Quick knowledge check

You want AI's help replying to an upset customer. What's the safe way to give it context?

APaste the customer's email exactly as it arrivedBInclude the customer's name and order history so the reply is accurateCDescribe the situation without names, account details, or anything identifyingDDon't use AI for customer replies at all

Checkpoint

03Section

Habit two: check before you rely

You know from What AI Can and Can't Do that confident and correct are unrelated. The starter habit that follows from it: before an AI output leaves your hands, give it one deliberate pass.

Ask: will anyone rely on this?: Brainstorming for your own eyes needs no ceremony. An email, a post, a report — those get the pass.
Verify anything factual: Names, numbers, dates, claims, links — confirm each in an original source before repeating it. Asking the AI "are you sure?" doesn't count.
Read it as the recipient: Does it say what you actually mean, in a voice you'd stand behind? Fix what doesn't sound like you.

Fact-check helper prompt

Checkpoint

04Section

Habit three: you own what goes out

Underneath the first two habits sits the rule that makes sense of both: the human stays responsible. When AI-drafted work goes out under your name, it's your work. Your client, your boss, your audience will hold you to it — "the AI wrote it" has never been an accepted excuse anywhere it matters.

The same rule sets a hard boundary: legal, medical, financial, and employment decisions need a qualified human, full stop. AI can help you understand a topic and prepare better questions — it doesn't replace the professional, and nothing in this lesson is professional advice.

Key takeaway

AI can draft it. Only you can be responsible for it.

Quick knowledge check

An AI-drafted email you sent to a client contained a wrong price. Who's responsible?

AThe AI company — their tool made the errorBYou — it went out under your name, which is why the double-check habit existsCNobody — everyone knows AI makes mistakesDThe client, for not verifying the price

Checkpoint

05Section

Your starter checklist — and the deeper path

The three habits, condensed into something you can keep next to your keyboard:

No passwords, account numbers, or login codes — ever
No other people's private details — describe the situation, don't paste it
Verify any fact, name, or number you'll repeat — in an original source
Read important outputs as the recipient before sending
Legal, medical, financial, employment calls go to a qualified human

That's everyday safety handled. When you're ready for the deeper layer — verification routines for high-stakes work, handling genuinely sensitive data, and the ethics questions — the AI Safety & Privacy path picks up exactly where this lesson stops.

AI Safety & Privacy path

Pause and think: Which of the three habits are you most likely to skip when you're in a hurry? That's the one to make automatic first.

Checkpoint

Prompt exercise

Build your personal safety checklist

The starter checklist above is generic — yours shouldn't be. Copy this prompt into ChatGPT, Claude, Gemini, Copilot, or whichever AI tool you have access to (the website doesn't run AI itself), answer its questions honestly, and keep the checklist it produces somewhere you'll see it.

Act as a practical AI-safety coach — calm and specific, never alarmist. Ask me one question at a time about how I actually use AI: what kinds of tasks I bring to it, what kinds of information pass through my work and personal life, and whose information besides my own I handle. After four or five questions, write me a personal "before I paste" checklist: five to seven short rules tailored to my situation, in plain language I could pin above my desk. One rule for our conversation: remind me at the start that I shouldn't share actual passwords, account numbers, or other people's private details with you while answering — descriptions are enough.

Reflection: Compare your personal checklist with the starter one from this lesson. What did the AI add that's specific to your situation? That difference is your real risk surface.

Quick check

3 quick questions — no pressure

There's no pass or fail here. Answer them all, and we'll show you the answers either way.