Skip to article

OWN IT. SECURE IT. PROTECT IT. MONITOR IT.

Start Here: Use AI Safely, the Easy Way

By

GoBauer LLC

Article date:

October 15, 2026

Last reviewed:

October 15, 2026

Plain steps for any adult who uses artificial intelligence (AI). Secure your email, calendar and chat first, because messages you didn't choose can try to give your AI orders.

Drug ads on TV tell you the side effects. "May cause dizziness. Ask your doctor." They do that so you can choose with your eyes open.

AI offers don't do that yet.

ChatGPT, Gemini, Claude and Copilot are AI assistants. So are Meta AI, Grok and others. An AI assistant is a computer program you can talk to. It can write, explain, sort and sum things up for you. Lately, these assistants keep offering to do more, like reading your email. "Want me to find emails you can unsubscribe from?" "I can clean up your mailbox."

AI can really help. We use it in our work. It's a force multiplier: it helps you write, learn and plan faster. But it multiplies mistakes and scams just as easily. This guide shows you how to get that help while sharing less private information and keeping a person in charge.

We secure email, calendar and chat first, because other people can put messages there. If you let AI read them, those messages may carry hidden instructions. The same risk can come from websites and shared files.

Then we cover your settings, sign-in, what you share and how to check the responses. At the end, a setup prompt has AI walk you through it all, one step at a time, in about 30 to 45 minutes.

Four steps you'll use again and again. Own IT, Secure IT, Protect IT and Monitor IT are a simple cycle for staying safe online with any account or device: know what you have, lock it down, share carefully, and check in regularly. This guide applies them to AI. Monitor IT sends you back to Own IT, so safety becomes a habit, not a one-time setup.

The warning label

⚠ WARNING LABEL: POSSIBLE SIDE EFFECTS OF GIVING AI ACCESS

Uses: Sorting mail, summing up long emails, writing replies, organizing files.

Possible side effects:

  • It may read more than the job needs. Depending on what you allow, it may read all your email and files. That includes bank notices, doctor reminders, family messages and login codes.
  • Email can try to give it orders. A scam email can hide words you can't see. Those words are written for the AI, not for you. Even a friend's email account can be hacked.
  • Your information can leave without you hitting Send. In some cases, a picture in an AI-written email can carry your information out, like a secret written on the outside of an envelope.
  • Its responses can be wrong. A phone number or link in an AI response or summary may come from a scammer.
  • Stopping isn't undoing. Use Stop or Cancel, then check that the work stopped. This cannot undo completed actions or erase what the service already received. Check what's kept, for how long, and how to delete it.
  • Installs can leave ongoing access. A plugin or add-in can keep reading after today's job is done.

The biggest risk: you don't control what goes in

An AI with broad inbox access may retrieve messages beyond the one you meant to discuss. You picked some of those emails. Others may have arrived without your choice: strangers, spam, scams, fake "security alerts," and messages from friends whose accounts were hacked. Ignored isn't deleted. A trick that never fooled you could be waiting for the AI.

Use AI on email you chose to show it, not on email that chose you.

Choosing the message yourself helps, but it doesn't make what's inside trustworthy. A friend's account can be hacked, and a web page or shared file can hide instructions too. That's why you clean it up first and check what comes back.

Why we suggest turning off the AI in your email, calendar and chat

Your email, calendar and chat are the places where other people, including strangers and hacked accounts, can put words in front of your AI without asking you. A calendar invitation is another way outside content can reach your schedule. When AI is built in, it reads those words for you as it sums up, translates or suggests replies.

You may not have turned it on. Email and device companies have been switching AI on by default. They usually announce it in a blog post, release notes or one email that's easy to miss, and rarely with anything like a warning label. In January 2026, Google announced Gemini features turned on in Gmail, starting in the U.S. In January 2025, Microsoft added Copilot to Microsoft 365 Personal and Family plans, along with a price increase. That same month, Apple began turning on Apple Intelligence by default on newer iPhones and Macs. Updates can also turn features back on after you turned them off. So check your settings, even if you never said yes, and check again each month.

This isn't just a theory. In September 2026, KnowBe4 showed Gemini in Gmail pulling two login codes from other emails into a simple translation. The trick was hidden text a person couldn't see, and it got past Gmail's filters. Gmail's other defenses kept those codes from being sent out, which shows that layers of protection help. But in a second test, a different setup built on Gemini did send a code to an outside server, while newer Gemini models resisted the trick. KnowBe4 also reported real scam emails aimed at AI assistants. In 2025, other researchers used hidden words in a calendar invite to get Gemini to take actions the user never asked for. Google has added defenses, and says the work is ongoing.

If you leave it on, here's what can happen:

  • A hidden message gets the AI to show or send a login code or private detail.
  • A summary or to-do looks normal but was planted by a scammer, like "call this number to fix your account."
  • A fake invite puts a bad link or phone number on your calendar.
  • The AI reads and acts on mail you never opened.
  • Depending on your settings, your email can feed AI in other products.

Your call. You know your own risk, and your family's. Pick the level that fits:

Cautious

What you do: Turn email, calendar and chat AI off. Keep spelling and grammar. Use AI only on text you copied and cleaned up.

Good fit if: You keep taxes, health or money records in email, or share an account with family

Balanced

What you do: Turn it off, then turn on only the feature you use. Never act on a code, link or phone number from an AI summary.

Good fit if: You use one feature a lot and check what it tells you

Accepting

What you do: Keep it on. Set images to "ask first," turn off dynamic email, keep invites from strangers off your calendar, and check the real email before you act.

Good fit if: Convenience matters most, and your inbox holds little that's private

Whatever you pick, you can change it later. Monitor IT means you look again next month.

If you only do one thing today, find the AI switch in your email and choose your level. The setup prompt in the next sections walks you through it, or Setup Steps lists where the switch is for Gmail, Outlook.com, Hotmail, MSN, Yahoo, AOL and Apple Mail, with dates, since menus change.

Three rules

1. Don't connect your email, files or other apps

Connect and Allow are buttons that hand AI a key to your email, files or other apps. Leave them alone for now. You can still use AI with less access.

Start with no keys. Give AI only the keys a task needs, and take them back when it's done.

Big companies apply the same rule to AI. In September 2026, Nvidia's CEO, Jensen Huang, told CNBC that no matter how smart an AI agent is, the first step is to take away all of its rights, then give it access to files, tools or the internet only if it needs them, the same way employees only get the access their job needs. At home, you're your own IT department. And many AI features arrive with the keys already handed over, so the setup prompt starts by taking them back.

"Sign in with Google" (or Apple or Microsoft) is different: it signs you in, but it doesn't by itself let the AI read your email. Read the screen before you click, though. If it also asks to see your email, files or contacts, that's a Connect in disguise.

Watch for AI that's already built in, too. Many email apps now have AI that's on by default, like Gemini in Gmail. The setup prompt helps you decide what to do about it. And some connections copy your data to another company: if you add a Gmail account to Microsoft's new Outlook app, Microsoft's help page says a copy of your email, calendar and contacts goes to Microsoft too.

Before you share, install or click Connect, ask:

  1. What private data can it reach?
  2. What untrusted content will it read?
  3. How could it send, share or change something?

If you can't answer, don't connect. A person stays in charge.

2. Be the person in the middle

Don't copy straight from an email or document into a chat, and don't drag in whole files. Copy just the text you need into a plain-text app like Notepad or TextEdit first. Remove names, account numbers, codes and private details. Read it once more. Then paste the cleaned-up text into the chat.

Why? Some apps start working on what you paste before you press Send. And files carry hidden details, like who wrote them and the file name itself. Plain text leaves those behind, so you see the words you're sending. It doesn't remove tricks hidden in the words themselves, so read what you paste, and check the response.

General rule: one task, one chat. Long chats drift. The AI can only keep so much in view, so your early instructions fade, and it starts forgetting rules, repeating itself or making things up. When that happens, or when the topic changes, start a new chat. Setup Steps shows how to have the AI write a handoff summary so you can pick up where you left off.

3. Find the Stop button first

Before AI does anything for you, find its Stop or Cancel button. Closing the window usually doesn't stop work the AI service already started, because the work runs on the company's computers, not yours.

Before you start: which way to sign in

When you create an AI account, you can usually sign in with your own email and password, or with Google, Apple, Microsoft or Facebook. Either way can be safe. What matters more than the button is:

  • Strong sign-in on the account you use: a passkey, or 2-step verification (MFA) with an authenticator app.
  • A way back in that you control: current recovery email and phone, and backup codes saved on paper.
  • What you said yes to: signing in alone doesn't let the AI read your email. A screen that asks for more, like your email, files or contacts, does.

Signing in with another account means one key opens both, which is simpler but ties them together. A separate email and password keeps them apart, which is the approach we use. The one choice to avoid is a reused password with no MFA. Setup Steps compares the options.

Let AI walk you through setup

  1. Open your AI assistant in a web browser, not an app built into your computer. A basic browser chat without device tools or extensions is a narrower starting point, though it still uses anything that's turned on in your AI account, like memory or connections. The setup checks those.
  2. Sign in, or create a free account. Use a sign-in you'll protect, as described above. Signing in lets you lock the account, turn off what you don't want, and delete what you shared by accident. It also keeps a history, which is why the setup has you check what's stored.
  3. Start a new chat with nothing connected.
  4. Open the Quick Setup Prompt, copy it, and paste it in. It has no private information in it.
  5. Answer one step at a time. Reply "done," "skip," "help" or "pause." If you can't find a setting, skip it; the AI tells you what risk is left and lists what's still open at the end. Skipping a step doesn't make it safe.
  6. Describe what you see in words. Don't paste screenshots or personal details. If the AI asks for a password or code, stop.

It secures your AI account before it asks about you. The first four tasks lock your AI account, turn off training, and check memory and ads, before any questions about your devices or email. Some free AI plans now show ads, and in ChatGPT, for example, ads can be personalized using your chats and memory unless you turn that off. Ads aren't advice, so don't click them during setup.

You'll always know where you are. Each message starts with a progress line, like "Task 9 of 18 | SECURE IT | about 20 minutes left." The whole setup takes about 30 to 45 minutes. Need a break? Reply "pause," and the AI gives you a short note so you can pick up later in a new chat.

The AI only gives instructions. It isn't watching your screen or checking your settings, so "done" means you did it, not that anything was confirmed. It can also be wrong about where a setting lives, because menus change often. The prompt tells it to say so instead of guessing.

What it covers: first, the AI account you're using; then the same four steps, in order.

  • FIRST (tasks 1 to 4): sign in, lock the account with a passkey or 2-step verification (backup codes saved on paper), turn off training, and check memory and ads.
  • OWN IT: a quick list of your devices, browsers, email and apps, and the AI built into each, and what you won't share. Choose the fast track to answer a fill-in list in one message, or go one question at a time. The AI turns your answers into an Own IT list and asks about things people often forget. Your Own IT list is a map of your setup, so only build it in a signed-in account with training and memory off, use major versions only (like Windows 11, not a build number), and keep the full list outside AI. If you stay signed out, the AI asks only general types instead. Anything extra goes on a "later" list.
  • SECURE IT: the AI in your email, calendar and chat first (at the level you choose). Then the AI built into your computers, phones and browsers, what's stored, web search, connections and the Stop button.
  • PROTECT IT: the person-in-the-middle habit, asking well, checking responses, and starting a new chat with a handoff summary.
  • MONITOR IT: what to watch for and a monthly checklist, which sends you back to Own IT.

It works with Gmail, Outlook.com, Hotmail, MSN, Yahoo, AOL and Apple Mail.

Rather do it yourself? The Setup Steps page has the same steps in the same order, with every detail.

AI at work: use the approved tools

Shadow AI means using AI tools at work that your company hasn't approved, like pasting work email into a personal chatbot. It's usually done to get work done faster, but it can put customer, patient or company information where it doesn't belong.

The answer isn't to avoid AI. Used well, it helps one person do the work of several, and the same multiplier works for mistakes. So:

  • Follow your company's AI rules. Use the tools it approved, with your work account.
  • Don't paste work information into a personal AI account.
  • Don't move work to your own phone or computer to get around the rules. Typing work information into a personal AI app, taking a photo of your screen, or forwarding work email to a personal account is still shadow AI, even if the company's controls can't see it. Once it's in a personal account, your company can't protect it or delete it.
  • Ask. If an AI tool would save you time, tell your supervisor or leadership. If the rules or approved tools get in the way of good work, say so. That's how better tools get approved.

If something seems wrong

  • Stop it. Use the app's Stop or Cancel button. Closing the window usually doesn't stop work the AI service already started.
  • Take away its access in the AI app and in your email account.
  • Look around. Check for sent emails, new rules that move or forward mail, and new devices.
  • Get help early. Ask someone you trust. At work, tell your IT team right away.

We take care of more than our own information. Our families, friends and coworkers trust us with theirs. We want every AI app to protect private information by design. Until it does, we have to look out for ourselves and each other.

Keep going

Own IT. Secure IT. Protect IT. was the 2019 Cybersecurity Awareness Month theme from CISA and the National Cyber Security Alliance. Monitor IT is the author's addition, because the first three are a theme and setup, but adding the fourth turns it into a repeatable lifecycle and sends you back to Own IT. This is an independent guide, not a CISA publication or endorsement. The author develops commercial security tools in this area; this guide is tool-neutral and recommends no product. Assistants and apps named here are examples, not endorsements. Safer does not mean risk-free. CC BY 4.0.

Sources

Explore the guides

All AI safety guides | Reference library

Back to all resources