Personal AI Security Checklist

Check how you use artificial intelligence (AI), one pillar at a time. Leave anything uncertain open and come back to it. For help with settings, use the Checkup Prompt. For help with tasks, use the Safer AI Recipes. New to security basics like passwords, scams and backups? Start with the Cybersecurity Basics Checkup at GoBauer.com.

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?

Before AI acts for you, know how to pull the plug (S12).

If you can't answer these questions and limit access, don't connect. Use selected, reviewed text in chat with connections, web search and agent actions off. Web search counts: it sends words from your conversation to a search company.

OWN IT: Know your AI tools, accounts and connections, what they store and who else gets a copy.

• ☐ O1 Start with the task. Know the result you want, how you'll check it, how you'd finish without AI, and how much risk you're willing to take.
• ☐ O2 Find every AI account. Search your inboxes for "welcome," "verify," "subscription" and "receipt." Check card statements and app store subscriptions.
• ☐ O3 Find the ones you didn't sign up for directly. Check "Sign in with Google, Apple or Microsoft" pages, browser extensions, phone and desktop apps, and AI built into email, office, keyboard and meeting tools.
• ☐ O4 Keep a private list. For each tool: sign-in method, plan, purpose and connections. Keep it on paper or in a private file outside any AI tool. Never write down passwords or codes.
• ☐ O5 Know what each tool can reach and do. Note what it can read, change or send, whether it runs scheduled tasks, and which plugins, skills or extensions it uses.
• ☐ O6 Find out who else gets a copy. Some connections copy your data to another company, under additional terms. Check the official help pages.

SECURE IT: Lock down sign-in, settings and access.

• ☐ S1 Use a passkey, or a unique password with multifactor authentication. Prefer an authenticator app over email or text codes.
• ☐ S2 Protect account recovery. Secure your recovery email and phone. Don't rely on email-only recovery where better options exist.
• ☐ S3 Never approve a sign-in you didn't start. Never share a code with someone who contacts you.
• ☐ S4 Update and lock your devices. Install updates for devices, browsers and apps. Lock your screen.
• ☐ S5 Turn off optional model training where offered.
• ☐ S6 Review saved memory. Read what the AI remembers about you. Delete what you don't want, then choose whether memory stays on. Some tools use memory to shape the searches they send out.
• ☐ S7 Disconnect what you don't need, starting with personal email and calendar. You can't control who writes to you, and old spam and phishing you ignored is still there for an AI to read. AI built into your email, calendar and chat counts too: choose your level, Cautious (off), Balanced (only what you use) or Accepting (on, with safety settings).
• ☐ S8 Turn off automatic searching for sensitive work. Look for settings that let the AI search the web, connected apps or saved files on its own. Search on purpose only when you need to, in a chat with nothing private in it.
• ☐ S9 Limit calendar invitations. Where supported, add invitations only from known senders or after you accept them.
• ☐ S10 Install only what you understand. Before you add a plugin, skill, extension, add-in or app, know what it is, who made it, why you need it, where it installs, what it can read and how to remove it. An install is a Connect button that stays; a document add-in may read document contents, depending on its permissions. Reviews and store listings don't prove safety.
• ☐ S11 Separate AI browsers. If an AI browser or extension can act for you, use a separate browser profile without your banking, email or password manager.
• ☐ S12 Know how to pull the plug. Before letting AI act for you, find its Stop or Cancel control, your kill switch, and try it on a harmless task. Know where to turn off scheduled tasks and remove its access. Stay in charge: approve anything before it's sent, shared, bought, deleted or changed.
• ☐ S13 Close accounts you no longer use, after saving anything you need. Unused accounts still hold your data.

PROTECT IT: Prepare, review and stay in control.

• ☐ P1 Confirm you have permission to share. Information about clients, coworkers, family or patients may not be yours to share. Follow work and school rules.
• ☐ P2 Share the minimum. Never share passwords, codes, recovery keys or full account numbers. Give health information extra care.
• ☐ P3 Clean up text before it reaches the chat box. Summarize it in your own words, redact what isn't needed, or swap sensitive details for consistent pseudonym labels like [PERSON_A]. Then review what's left, including the context. Typed text may be sent before you press Send.
• ☐ P4 Keep your label key private. The list of who [PERSON_A] really is, which you use to restore real names later, belongs outside AI.
• ☐ P5 Remember that labels don't make data anonymous. Context left behind, like a job title, a date or a town, can still point to a person.
• ☐ P6 Prefer plain text to whole files. Copy just the text you need, as plain text, and review it. If you must upload a file or image, check it for hidden comments, tracked changes, hidden sheets, author names and photo locations.
• ☐ P7 Use a temporary or private chat for sensitive one-offs where offered, and check what it still keeps.
• ☐ P8 Treat outside content as untrusted, even from familiar senders. Emails, invitations and webpages can hide instructions meant for an AI. Content you ask an AI to examine should not become permission for it to act.
• ☐ P9 Check before you rely on it. Compare facts, totals and dates with the originals and their sources. Don't open links in AI answers just to see where they go.
• ☐ P10 Act yourself. Send, buy, cancel, delete and share from the official app or website, not through an AI acting for you.
• ☐ P11 Stay with AI that acts for you. Treat an AI agent like cooking on a stove: stay nearby. Before you walk away, use Stop or Cancel and confirm it stopped.
• ☐ P12 Keep sensitive chats private. Don't make public share links for chats with personal details. Shared chats have turned up in search results.
• ☐ P13 Use AI on email you chose to show it, not on email that chose you. Copy the message you picked into a separate chat with nothing connected. Keep built-in email AI only at the level you chose (S7). Skip messages from senders you don't know. Treat links, phone numbers and "security alerts" in an AI summary as unverified until you find them yourself in the official app or on a statement.
• ☐ P14 Close out each task. Save what you need on your own device, then delete the chats, uploads and prepared copies you no longer need.
• ☐ P15 One task, one chat. When the AI drifts, start fresh with a reviewed handoff summary.

MONITOR IT: Check monthly and after updates.

• ☐ M1 Look for new permissions and connections you didn't add, new AI features you never turned on, and settings an update switched back on.
• ☐ M2 Look for surprise changes: unfamiliar memories, inbox rules, forwarding, calendar events, shared files or sent messages.
• ☐ M3 Watch account alerts. Turn on alerts for new sign-ins, new devices and apps given access, and compare them, plus password resets and receipts, against your list. Check each alert by signing in yourself, not through the email's link. Sign up for breach alerts for your email addresses.
• ☐ M4 Sweep for leftovers. Delete old chats, uploads, memories and share links that cleanup missed, after saving what you must keep.
• ☐ M5 Search for yourself. Look for public AI share links or files you didn't mean to publish.
• ☐ M6 Retire what you're done with, then start again. Each quarter, revoke unused access at the source, update your list, recheck settings and return to Own IT.

If something goes wrong

• You shared something you shouldn't have. Stop adding more. Delete it with the tool's controls. Report it if it involves work, school or someone else's information.
• An account, password or code is exposed. From a trusted device, change it, sign out other sessions and check recovery settings.
• An AI agent did something unexpected. (1) Use the app's Stop or Cancel control; typing "stop" or closing the window isn't enough. (2) Turn off scheduled or repeat tasks. (3) Remove its access in the AI app and at the source account. (4) Check those accounts for messages sent, files shared, events created or charges. (5) Check that it really stopped, including scheduled or queued work. (6) Undo what you can; pulling the plug can't take back what already happened.
• An AI summary or answer tells you to call, click or act fast. Don't use the number or link it gave you. Find the real one in the official app, a trusted bookmark or a statement.

Keep the right expectations

Training, memory and deletion are three different switches, and signing out isn't a privacy guarantee. Turning things off reduces exposure but doesn't guarantee containment, and it doesn't erase data already received. A checked box records a decision, not a guarantee of safety. Until AI services protect our data by design, we're the stewards: learn the controls, or connect nothing and use reviewed text in a basic chat.

---

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. Independent guide; not a CISA publication or endorsement. The author develops commercial security tools in this area; this checklist is tool-neutral and recommends no product. CC BY 4.0. Provided as is, without warranty.

Explore the guides

All AI safety guides | Reference library