Millions of people now treat ChatGPT like a diary, a therapist, a lawyer, and a coworker all in one chat box. That’s exactly the problem.
Every message you send isn’t a private conversation vanishing into thin air. Depending on your settings, it can be stored, reviewed by human trainers, used to improve future models, or exposed in a data breach. In 2023, a bug in ChatGPT briefly let some users see snippets of other people’s chat titles and payment details — a reminder that no chatbot is breach-proof.
Read Also: 5 Things Claude Can Do Right Now That Will Save You 2 Hours Today
This isn’t about fearing AI. It’s about using it like a smart tool, not a confessional. Below are the 10 things you should never share with ChatGPT, why each one is risky, and what to do instead.
1. Your Social Security Number or National ID
Never paste your SSN, BVN, NIN, passport number, or any government ID into ChatGPT — even if you’re just asking it to “check if this looks like a valid format” or draft a form. These identifiers are the backbone of identity theft. If a breach or misconfiguration ever exposes chat logs, a leaked ID number is far more damaging than a leaked email address.
Do this instead: Ask ChatGPT to explain the format or requirements of an ID document without giving your actual number.
2. Passwords and Login Credentials
It sounds obvious, but people do this constantly — pasting a password into ChatGPT to ask “is this strong enough?” or copying an entire error log that happens to include a saved login. Chat histories aren’t designed with the same security architecture as a password manager.
Do this instead: Use a dedicated password strength checker or password manager (like Bitwarden or 1Password) instead of an AI chatbot.
3. Banking Details and Financial Account Numbers
Account numbers, card numbers, CVVs, and banking app screenshots should never go into a chatbot, even to “help me understand this transaction” or draft a dispute letter. If you need help with a financial document, redact the account numbers first.
Do this instead: Black out or replace real numbers with placeholders (e.g., “Account: XXXX1234”) before sharing any financial text with ChatGPT.
4. Medical Records and Health Diagnoses
Asking ChatGPT general health questions is fine. Uploading your actual lab results, a scanned prescription, or a doctor’s letter with your name and diagnosis is different — that’s protected health information in most countries, and AI platforms generally aren’t built or certified to handle it the way a hospital’s system is.
Read Also: How to Write Your First Claude Prompt (And Get a Useful Answer Every Time)
Do this instead: Describe your symptoms or question in general terms without attaching documents that contain your name, date of birth, or medical record number.
5. Confidential Work Documents and Trade Secrets
This is one of the most common — and most career-damaging — mistakes. Employees have pasted internal source code, unreleased product specs, client contracts, and strategy decks into ChatGPT to “summarize this” or “fix this bug.” Several major companies, including Samsung, restricted employee ChatGPT use after internal code was reportedly shared this way. Once proprietary information leaves your company’s systems, you generally can’t guarantee it stays contained.
Do this instead: Use your company’s approved AI tools (many enterprises now have private, contractually-protected versions of AI models), or strip out identifying details, real client names, and proprietary logic before asking general questions.
6. Someone Else’s Personal Information
Sharing a friend’s phone number, a client’s home address, or a colleague’s private messages to get advice puts someone else’s privacy at risk without their consent — even though the conversation feels private to you.
Do this instead: Anonymize names and details (“a colleague” instead of a real name) when asking for advice about a situation involving other people.
7. Explicit Plans to Break the Law
Beyond the obvious ethical issues, detailing real plans for fraud, hacking, or other illegal activity — even “hypothetically” — creates a written record. Most AI platforms log conversations, and that record can resurface in ways you don’t control.
Do this instead: If you’re researching a legal gray area for legitimate reasons (journalism, compliance, fiction writing), frame it clearly as research or creative work, not as a personal plan.
8. Your Exact Location or Travel Plans
Telling ChatGPT “I’m home alone at [address] until Friday” or detailing when your house will be empty for a trip is the digital equivalent of posting it on a public bulletin board — especially since you don’t fully control where that data travels or how long it’s retained.
Do this instead: Ask about a destination, itinerary ideas, or packing lists without attaching your home address or exact travel dates.
9. Your Full Legal Name Combined With Sensitive Context
It’s not any single detail that’s dangerous — it’s the combination. Sharing your full name alongside your financial struggles, immigration status, health condition, or workplace conflict creates a profile that’s far more identifiable and exploitable than either piece alone.
Do this instead: Keep sensitive conversations anonymous. Use “I” and describe the situation without your full name attached, especially in shared or exported chats.
10. Access Tokens, API Keys, and Two-Factor Codes
Developers sometimes paste an entire .env file or error log to debug faster — codes and keys included. An exposed API key or 2FA code can let someone else access your accounts, run up cloud bills in your name, or impersonate you entirely.
Do this instead: Redact keys and tokens (replace with sk-xxxxxxxx) before sharing code snippets, and always rotate any key that was ever pasted into a chatbot, just in case.
The Bigger Picture: Why This Matters
AI chatbots like ChatGPT are trained to be helpful, not to protect you from yourself. The responsibility for what you share sits with you. A simple rule of thumb: if you wouldn’t post it on a public forum under your real name, don’t paste it into ChatGPT.
This doesn’t mean avoiding AI — it means using it the way professionals use any powerful tool: with guardrails. Turn off chat history and training in your settings if you want extra privacy, use enterprise or team plans with stricter data policies for work tasks, and get comfortable anonymizing sensitive details before you hit send.
Frequently Asked Questions
Does ChatGPT save everything I type? By default, conversations may be stored and can be used to improve the model unless you turn off chat history or use a version with data controls (like ChatGPT Team, Enterprise, or the “Temporary Chat” feature). Settings differ by plan, so check your account’s data controls directly.
Can OpenAI employees read my chats? OpenAI states that authorized personnel may review conversations for safety and abuse monitoring purposes, particularly conversations flagged by automated systems. Assume anything you type could theoretically be seen by a human reviewer.
Read Also: What Is Claude AI? The Honest Beginner’s Guide for African Professionals
Is it safe to use ChatGPT for work? It depends on your employer’s policy and the plan you’re using. Many companies now offer approved, contractually-protected AI tools specifically so employees don’t need to paste confidential data into consumer-grade chatbots.
What should I do if I already shared sensitive information with ChatGPT? Change any exposed passwords or API keys immediately, monitor accounts tied to any shared financial or ID information, and consider deleting the relevant chat from your history going forward.
Written by Olasunkanmi Adeniyi, founder of AI Discoveries — practical AI education for professionals across Nigeria, Ghana, Kenya, South Africa and Rest of The World.
Leave a Reply