Privacy, Safety, and Responsible Use

A youth soccer coach sat down the night before the season opener to get organized. She had a spreadsheet with every player on it: first and last names, birthdates, home addresses, a parent’s cell phone for each kid, and a note next to two names about allergies and a heart condition. She opened her AI tool, pasted the whole roster in, and typed “turn this into a clean contact sheet and a phone tree.” A tidy result came back in seconds. It felt like magic. She had just handed a company a file of children’s names, addresses, phone numbers, and medical details, and she had no idea where that file would live afterward or who might ever see it.

She didn’t do anything unusual. This is one of the most common ways ordinary, careful people expose information they’d never post publicly. The tool is helpful, the task is boring, and the sensitive part slides by because you’re focused on getting the chore done. This chapter is about the small habits that keep that from happening to you and the people who trust you.

Why this matters to you

When you type into an AI tool, it doesn’t feel like publishing. It feels private, like talking to yourself or texting one trusted friend. That feeling is the trap. What you type travels to a company’s computers to be answered, and depending on your settings and your plan, it may be stored and may be used to help improve future versions of the tool.

That’s a reasonable tradeoff for a lot of what you do. Drafting a birthday message, rewording an email, explaining a math concept, planning a trip: none of that needs to be secret. The problem starts when the same easy habit gets pointed at information that isn’t yours to share, or that would hurt you if it leaked. Many people in this course are coaches, teachers, volunteers, and parents who routinely hold other people’s information, including children’s. When you’re the keeper of someone else’s data, the standard is higher, and the law may be too.

The goal here isn’t fear. Millions of people use these tools every day without incident. The goal is a simple filter you run before you paste, so the useful stuff stays useful and the sensitive stuff never leaves your hands.

What happens to what you type

Picture the difference between writing in a locked diary and mailing a postcard. A diary stays with you. A postcard passes through other hands on its way, and anyone handling it could read it. Typing into an AI tool is closer to the postcard. Your words leave your device, travel to a company, get processed there, and often get kept for some period of time.

Two things can happen to what’s kept. It may be stored in your account history, which is convenient for you and also means it exists somewhere you don’t control. And on many consumer plans, your conversations may be reviewed or used to help train future versions of the tool, unless you’ve turned that off.

The safe mental model is simple: treat anything you paste as if it could someday be seen by a stranger. Not because it definitely will, but because you can’t guarantee it won’t, and that assumption makes your decisions easy. If a sentence would embarrass you or harm someone on a postcard, it doesn’t go in.

Try It Now Write your personal “never paste” list. Open a note on your phone and finish this line five times: “I will never paste ______ into an AI tool.” Include the obvious ones below, then add two that are specific to your life, like a client’s file number, your child’s school login, or a teammate’s home address. Keep the note where you’ll see it. What did you notice? Most people find that once the list exists, the borderline cases get obvious fast. The hard part was never knowing the rule. It was deciding it before the moment you’re in a hurry.

What you should never paste

Some categories don’t belong in an AI tool, full stop. Passwords and login codes. Full financial account numbers, card numbers, and banking details. Government IDs like a Social Security or passport number. Private medical details, whether they’re yours or someone else’s. And the big one people forget: other people’s personal information, especially a minor’s. A child’s full name, birthdate, address, photo, school, or health note deserves more protection than your own, not less, and in many settings you’re legally responsible for guarding it.

Here’s the good news: you can usually get the tool’s help without handing over the sensitive parts. The move is to strip or fake the identifying details first, the same skill you practiced back in Chapter 6. If you want help writing a phone tree, you don’t need real numbers to design the format; do the layout with fake names and placeholders, then fill in the real details yourself afterward, in your own document. If you want advice on a hard message to a parent, replace names with “the player” and “the parent” and drop the address and phone entirely. The tool can shape the words. It doesn’t need to know who they’re about.

Back to the coach from the opening. All she needed was a clean layout and a phone-tree structure, and none of that requires a single real name. She could have pasted “Player 1, Parent 1, 555-0000” as placeholders, gotten the same tidy format, and typed the real roster into her own private file. Same result, none of the exposure.

Try It Now Take one real task you’ve been meaning to do that involves other people’s details, like a sign-up sheet, a contact list, or a note about someone’s situation. Before you use your AI tool, rewrite it with every identifying detail removed or faked: names become “Person A,” numbers become 555-0000, addresses disappear. Now run the task on the stripped version. What did you notice? For most people the tool’s answer is just as good, because the help you wanted was about structure and wording, not about the private facts. That’s the whole trick: separate the shape of the task from the sensitive contents.

Safer defaults, scams, and being a responsible user

You can lower your risk before you type a single thing by changing a few settings. Most AI tools let you do two useful things: turn off or limit whether your conversations are used to train the tool, and turn off or clear your chat history. Turning these on where available means less of what you type is kept and reused. The exact switches, and where they hide, change often and differ by tool, so this course keeps them in one place. See Module M1 (Meet the Current Models) for the current step-by-step on finding and changing your privacy and training settings in the tool you use.

The other half of responsible use is defending against AI-enabled tricks. The same technology that writes a warm birthday note can write a flawless scam email with no spelling mistakes, and can now clone a voice or a face from a short sample. People have received phone calls that sound exactly like their child, their boss, or a grandchild, begging for money or a code, generated by software from a few seconds of audio. The defense is not technical. It’s a rule: any unexpected, urgent request for money, codes, or secrecy gets verified through a second channel before you act. Hang up and call the person back on their known number. Text the account you already have. A real emergency survives a two-minute check. A scam usually doesn’t.

One more piece of responsible use is knowing that AI outputs can be biased. These tools learn from enormous amounts of human writing, and human writing carries human stereotypes, so the tool can quietly reflect them: assuming a nurse is a woman and an engineer is a man, or describing some groups in narrower terms than others. It’s not malice and it’s not always obvious, which is exactly why it’s worth watching for, especially when an output describes people.

Try It Now Run a deepfake gut-check. Read this scenario and decide your move before reading on: you get a voice message that sounds just like your teenager, crying, saying they’re stranded, phone dying, and need you to send money to a number right now and not tell the other parent. It sounds real. What do you do? What did you notice? The right move is to stop and verify through a second channel: call your kid’s actual phone, text them, or reach someone who’s with them. The urgency and the “don’t tell anyone” are the tells. A voice sounding right proves nothing now, and building the two-minute-check reflex today is what protects you on the day it’s your turn.

What people get wrong here

The biggest mistake is treating this as all-or-nothing: either AI is a privacy nightmare you should avoid, or it’s fine and you can paste anything. Neither is true, and both get people into trouble. The realistic stance is a tradeoff you make on purpose. You accept that most of what you type is low-stakes and the convenience is worth it, and you draw a firm line around the handful of things that aren’t. Reasonable, careful people land in slightly different places on where that line sits. What matters is that you drew one at all, rather than deciding fresh in every rushed moment.

The environmental question works the same way, and it’s worth naming honestly because you’ll hear strong claims on both sides. Running AI tools uses real electricity and water for cooling, at data centers that consume meaningful amounts of both, and that footprint is a genuine concern to many thoughtful people. It’s also true that estimates of how large the impact is, and how it compares to everyday things like streaming video or driving, vary a great deal depending on who’s measuring and what they count. Reasonable people weigh it differently, some using these tools freely and others sparingly. This course won’t tell you where to land. It will tell you the disagreement is real, the concern is legitimate, and being informed about it is part of using any powerful tool responsibly.

Your move

Do the two-minute privacy setup this week. Open your AI tool and, using the steps in Module M1 for your specific tool, find your training and history settings and decide, on purpose, how you want each one set. While you’re there, pull up the “never paste” note you started earlier and read it once. Then, the next time you’re about to paste something that involves another person, especially a child, pause and ask one question: does the tool actually need this real detail, or just the shape of the task? Nine times out of ten, the shape is all it needs, and the private part stays with you.

This chapter was developed with AI assistance and reviewed by a human editor. It’s educational, fact-checked where applicable, and may contain minor errors. It’s not a substitute for professional advice.