10 Things You Should Never Share With an AI Chatbot
AI chatbots can help with almost anything, but convenience can make a prompt box feel more private than it is. Here are 10 things you should keep out of consumer AI chats, plus safer ways to get useful answers.
Maya Chen | Future Tense
AI chatbots are easy to talk to. That is part of their appeal, and part of the risk.
A chat window can feel like a private conversation with a patient, helpful assistant. In reality, you are sending information to a technology service with its own systems, policies, retention rules and security controls. Those details vary by company and account type.
This does not mean you should stop using AI. It means you should use it with the same caution you would bring to email, cloud storage or an online form. Before you paste, upload or dictate anything, ask one question: Would I be comfortable if this information were stored, reviewed during a security investigation or exposed through an account breach?
If the answer is no, keep it out of the chat.
1. Passwords, PINs and security codes
Never give a chatbot a password, debit-card PIN, recovery phrase or one-time authentication code. AI does not need your actual credentials to help you create a stronger password policy or troubleshoot a login. Use placeholders instead, such as “MY_PASSWORD,” and store real credentials in a reputable password manager.
2. Social Security and government ID numbers
Your Social Security number, passport number, driver’s license number and tax identification numbers can be used to verify your identity. That makes them especially valuable to criminals. If you want help understanding a government form, remove the number, your full birth date and any barcode before uploading it.
3. Bank and payment information
Do not paste complete credit-card numbers, bank-account numbers, routing details or payment-app credentials into a consumer chatbot. You can ask AI to explain a charge, compare loan terms or build a budget using rounded figures and fictional account labels.
4. Complete, identifiable medical records
AI can be useful for preparing questions for a doctor or translating medical language. It is still wise to remove your name, date of birth, patient number, address and other identifying details. A diagnosis without a name may be useful context. A full record containing your identity, insurance information and medical history is far more than most prompts require.
5. Confidential workplace information
Source code, unreleased financial results, product plans, internal emails and trade secrets should not go into a public consumer chatbot. Your employer may provide an approved business AI service with different contractual protections. If so, follow the company’s policy and make sure you are using the approved account, not a personal one. The Federal Trade Commission has specifically warned that customers may reveal internal documents and their own users’ data to AI services.
6. Customer, client, patient or student data
Someone else’s personal information is not yours to give away. Names, contact details, case notes, school records and customer lists may also be protected by contracts, professional duties or privacy law. Ask the chatbot to work with a fictional example or a properly anonymized summary instead.
7. Private information about other people
The same rule applies outside work. Do not upload a partner’s private messages, a child’s records or a friend’s personal history simply because you want advice. Remove names and details that could identify the person. Consent matters, even when the technology makes sharing effortless.
8. Privileged legal documents or strategy
A chatbot can explain general legal vocabulary, but uploading attorney communications, settlement strategy or confidential case files may create privacy and privilege concerns. Use a licensed attorney for advice about your specific situation. If you use AI to organize questions, strip out names, case numbers and strategic details first.
9. Intimate photos and private media
Do not upload intimate images, private videos or sensitive voice recordings to a general AI service. That includes material involving another person, especially if they have not clearly consented. Once a file leaves your device, you are relying on the provider’s storage, access and deletion practices.
10. Your exact location and security routine
Avoid combining your home address, live location, travel dates, alarm details and daily schedule in one conversation. Each fact may seem harmless alone, but together they can create a useful map of when and where you are vulnerable. Ask for a travel itinerary using a city or neighborhood, then add exact addresses privately.
Privacy controls help, but they are not permission slips
Major chatbot providers offer settings that can limit how conversations are used. OpenAI, for example, lets users turn off “Improve the model for everyone”, while Google explains how Gemini activity is stored and used in its Privacy Hub. Those controls are worth reviewing, but policies differ and can change.
The safer habit is data minimization: share only what the task requires. Replace names with roles, remove account numbers, summarize long documents and use made-up examples whenever possible. Temporary or private chat modes can reduce some exposure, but they do not make a secret necessary to the prompt.
AI can help you think, write and plan without receiving the keys to your identity. As we have noted before, the real question is not whether AI is useful. It is how many keys to your life you are willing to hand over.
Treat the prompt box as a tool, not a vault. The best information to protect is the information you never send.
