Never share passwords, Social Security numbers, bank account details, or credit card numbers with any AI chatbot. Your conversations may be stored, reviewed by employees, or exposed in a data breach. Keeping sensitive details out of your chats is the single easiest way to protect yourself.
AI chatbots are genuinely useful tools. But they are not private diaries. What you type goes to a company's servers — and it can be stored, reviewed, or potentially exposed. Before you share, it helps to know where the line is.
Passwords and Login Credentials
Never type a password into a chatbot, not even to ask "is this password strong enough?" AI companies often store conversations to improve their systems, and employees may review those conversations for quality or safety checks. If the company ever suffers a data breach, your credentials go with it.
The fix is simple: describe the problem without the actual password. Instead of pasting it in, say "I have an 8-character password with a mix of letters and numbers — is that strong enough?" You will get the same advice with zero risk.
Social Security or National ID Numbers
Your Social Security number is the master key to your identity. One exposure can lead to fraudulent tax returns, new credit accounts opened in your name, and years of cleanup work. No legitimate AI task — writing an email, planning a trip, summarizing a document — ever needs your SSN. If you are filling out a form that requires it, do that on the official website, not in a chatbot.
Bank Account or Credit Card Numbers
Chatbots cannot move money or process payments. There is no reason to type your bank account number, routing number, or card number into one. Beyond the storage risk, phishing apps disguised as AI tools are specifically built to harvest this kind of information. Type those numbers only into verified, official bank or payment websites.
Detailed Medical Records or Diagnoses
It is perfectly fine to ask a chatbot general health questions: "What does 'LDL cholesterol' mean?" or "What are common symptoms of a sinus infection?" But sharing your actual medical records, specific diagnoses, or prescription details is a different matter. If that data is ever breached, it could theoretically be used against you by insurers or employers in certain industries. Use general language instead — the AI can still give you useful information without knowing it is about you.
Your Children's Full Names, Schools, or Photos
Children's information is especially sensitive, and federal privacy law (COPPA) exists for good reason. AI chatbots are not designed to hold your child's personal details, and if a conversation were ever exposed, information about where your child goes to school or their daily routine could create real safety risks. Use first names only, or just say "my daughter" or "a 10-year-old." The AI will still help you with homework questions, parenting advice, or anything else you need.
Legal Matters You Are Involved In
Attorney-client privilege does not extend to AI chatbots. If you are in a lawsuit, a custody dispute, a divorce, or any legal proceeding, do not type the details into a chatbot. That conversation could potentially be subpoenaed or introduced as evidence. Ask your attorney what you can and cannot discuss online, and treat AI chat the same as a public forum — because for legal purposes, it effectively is one.
Trade Secrets or Confidential Work
Many companies have already banned employees from pasting internal documents into public AI tools, and for good reason. Confidential client data, unreleased product plans, internal financial figures — these belong to your employer or client, not a tech company's database. Check your company's policy before using AI for work tasks. If you are unsure, describe the situation without naming the client or sharing any actual documents.
Your Home Address Combined With Your Daily Schedule
A street address on its own is usually low risk. But combining your precise home address with details about when you leave for work, when your house is empty, or your daily routine creates a different kind of exposure. If chatbot logs were ever leaked, that combination of data is exactly what a bad actor needs. Keep location details vague — "a suburb of Chicago" or "a neighborhood near downtown" is almost always enough context for the AI to help you.
What to Try Next
To see exactly what a chatbot service already knows about you and what it keeps, read what AI chatbots actually know about you. Then take five minutes to walk through how to adjust your ChatGPT privacy settings — a few simple toggles can significantly reduce how much gets stored.
