Think of an AI chatbot like a helpful stranger on a train: great conversation, but you wouldn’t hand them your wallet. Here’s the definitive list of what you should never tell an AI chatbot — and the reasoning behind each item, so the rule actually sticks.
What You Should Never Tell an AI Chatbot: The Never-Share List
1. Passwords and Login Credentials
This should be obvious, but people do it — pasting passwords when asking for account help, sharing API keys when debugging code. Once typed, they’re in a log on someone else’s server. Rotate any credential you’ve ever pasted into a chatbot. When you need AI help with code, use redacted examples.
2. Financial Details
Bank account numbers, credit card numbers, investment balances, crypto wallet keys. “Help me understand my bank statement” is fine after redacting account numbers. The AI doesn’t need your real numbers to explain the concepts.
3. Government ID Numbers
Social Security numbers, national IDs, passport numbers, driver’s license numbers. There is essentially no legitimate reason a chatbot needs these. If a task seems to require one, rethink the task.
4. Precise Health Information
General health questions (“what causes migraines?”) are fine. Your actual diagnosis, medications with dosages, lab results with your name on them — not fine. Beyond privacy, there’s a safety angle: chatbots aren’t doctors, and detailed personal health data plus confident-sounding AI advice is a risky mix.
5. Legal Troubles and Strategy
Details of lawsuits, disputes, or legal strategy. Beyond the privacy risk, sharing case details with third-party services can complicate things legally in some jurisdictions. Talk to your lawyer about specifics, not a chatbot.
6. Children’s Personal Information
Kids’ full names, schools, photos, routines, medical details. Children can’t consent to having their data logged, and this data persists. Be especially careful with homework help — “my son Jake’s essay” should become “a student’s essay.”
7. Proprietary Business Data
Unreleased products, trade secrets, internal financials, customer lists, proprietary code. This is how companies leak — one well-meaning employee pasting a confidential doc “to summarize it.” Many employers now ban this explicitly; check your company’s AI policy.
8. Intimate Personal Details
Relationship specifics, sexual content involving real people, deeply personal confessions. Even with the best privacy settings, you’re creating a permanent record on someone else’s infrastructure. A journal (paper or local app) is the right place for this.
9. Location and Routine Details
Your home address, daily schedule, travel plans with dates. “I’ll be away from home July 3-10” combined with your address is a burglary invitation if data ever leaks. Keep itineraries vague.
10. Anything Covered by an NDA
If you signed a non-disclosure agreement, assume pasting covered material into a chatbot violates it. When in doubt, don’t.
Why “But I Deleted It” Isn’t Enough
Deleting a chat removes it from your view, but providers retain data briefly for safety and abuse processing, and anything used for training (if you haven’t opted out) can’t be un-learned from a model. Deletion is hygiene, not a time machine. The protection has to happen before you hit send.
The Anonymization Habit
You don’t have to avoid chatbots for sensitive topics — just strip the identifiers first:
- “My daughter Emma, 14, at Lincoln High” → “a 14-year-old student”
- Paste of your actual contract → paste with names, amounts, and dates replaced by [X]
- Your company’s proprietary code → a minimal reproduction of the bug with fake variable names
You get 95% of the AI’s helpfulness with ~0% of the risk. For the nuclear option — sensitive work with zero exposure — run a model locally.
Teaching Kids and Family These Rules
You’re careful — but is your family? Kids and less tech-savvy relatives are the weak links:
- For kids: make it a simple rule — “never tell the computer your real name, school, or address.” Repeat until automatic. See are AI chatbots safe for kids for the full guide.
- For parents/grandparents: they’re prime targets for AI-assisted scams and oversharing. A 10-minute conversation about “the chatbot is not your friend, don’t tell it bank details” prevents real harm.
- For employees: if you run a team, this isn’t optional — make it policy. Most corporate AI data leaks are well-meaning staff pasting the wrong thing.
The “Pause Before Paste” Habit
The single most effective protection: pause before pasting anything substantial into a chatbot. Three seconds, one question: “Does this contain anything from my never-share list?” That micro-habit catches 95% of mistakes — the accidental API key in a code paste, the real names in a “hypothetical” scenario, the account numbers in a financial question. Make it as automatic as checking your mirrors before driving.
What If You Already Shared Something?
Don’t panic — act:
- Delete the conversation from your chat history immediately.
- Check training opt-outs — if you were opted in, your data may be in the training pipeline. Opt out now to prevent future inclusion.
- Rotate compromised credentials: any password, API key, or token you’ve pasted should be changed. Treat it as exposed.
- Monitor: for financial data, watch statements for unusual activity for a few months.
- Learn: figure out why it happened (rushing? didn’t know?) and fix the process, not just the incident.
One mistake doesn’t doom you — providers process billions of chats, and a single deleted conversation is a needle in a haystack. But treat it as the wake-up call to build the habits above.
OpenAI’s ChatGPT remains the baseline most chatbots are measured against — useful context for any comparison.
Bottom Line
What should you never tell an AI chatbot? Passwords, financial data, ID numbers, detailed health info, legal specifics, kids’ data, business secrets, intimate details, location routines, and anything under NDA. Memorize the categories, build the anonymization habit, and you’ll use AI freely without the low-grade anxiety. For the broader picture, see whether it’s safe to share personal info with chatbots.
