AI chatbots such as ChatGPT, Grok, Claude, and Gemini will be commonplace tools for anything from idea generation to document summarisation and code generation. They are quite beneficial.
Conversations you have with these models are not as private or secure as you would think. Unless you opt out, major AI businesses’ privacy rules typically permit them to collect chat data to train their models. Data may be exposed to a breach or examined by humans. Everything you enter may be saved, examined, and combined with other information about you, as Stanford researchers have noted.
8 Specific Things to Not Share With Your AI Chatbot
Pro Tip: Always keep your personal data offline, you’re safe
1. Passwords and usernames (or any login credentials)
Never enter recovery codes, usernames, passwords, API keys, or documents that contain them into a chatbot. Sensitive information may be stored or disclosed even if you are debugging an error message.
Additionally, AI tools are terrible at creating secure passwords. Use passkeys or stick with a trustworthy password manager.
2. Financial Data
Don’t provide any information about your personal finances, such as bank statements, credit card numbers, account balances, investment information, routing numbers, or tax records.
AI chatbots are not licensed financial advisors. If you disclose personal information, you run the risk of being targeted by scammers, fraud, or identity theft. Financial questions should be kept broad (e.g., “Explain how a Roth IRA works” rather than “Review my specific portfolio statement”).
3. Medical Records
Don’t upload any medical documents, lab results, diagnoses, or prescription information. AI is not a medical professional. It is not appropriate to utilize it for medical advice.
In addition to erroneous answers, your personal health information may end up in training datasets or be compromised in a data breach. Insurance or future medical care may be impacted by this.
4. Personally Identifiable Information (PII)
This includes any information that might be used to identify you specifically, such as your full name, home address, phone number, email address, date of birth, Social Security number, passport number, and driver’s license details.
Identity theft is much facilitated when PII is combined with other information. AI prompts should be handled as public posts. When you can, make generalizations.
5. Overall Health Data
Details that don’t seem dangerous can be dangerous. For instance, requesting “heart-friendly dinner recipes” or disclosing information about drugs, sexual health, or certain ailments could enable the system (or consumers of the data downstream) to deduce sensitive health profiles. Advertisers, insurance companies, and other parties may receive these. Health-related prompts should be generic and anonymous.
6. Mental Health Issues
Artificial intelligence chatbots are not therapists. Even if some models have better safeguards, they can nevertheless respond to a crisis in an ineffective, generic, or even dangerous way. They should never be used in place of expert mental health assistance. Contact a crisis hotline or a licensed human counselor if you’re having trouble.
7. Personal Photos
There are several risks associated with uploading images, particularly those of yourself, your family, or your kids. GPS position information is frequently included in image metadata. Training datasets can be expanded to include the images themselves.
Strip EXIF data first, at the very least. Steer clear of pictures of children or anything really private. Think about whether the advantages of AI image editing are truly worthwhile.
8. Company or Confidential Work Document
Work-related uploads, such as internal reports, source code, client data, unreleased strategies, NDAs, or anything marked secret, should be handled with extreme caution.
Feeding private company data into public AI tools is expressly forbidden by many employers. Even summarising a paper may unintentionally reveal confidential information. When available, use approved private AI instances or make high-level inquiries. Additionally, to protect your AI account and its chat history from unauthorized access, always ensure your accounts are secured with Two-Factor Authentication.
Final Tips for Safer AI Use
Presume that nothing is confidential. Consider each prompt as possibly permanent and public.
Choose not to participate in training. Examine your AI tools’ settings (most now offer a way to turn off using your chats for model training).
Make generalizations about everything. Say “Explain common causes of authentication errors in Python” rather than “Help me fix this error in my specific bank login code.” Examine the policies. Different suppliers have different privacy policies. For the tools you use most frequently, read the fine print.
When in doubt, don’t include it. Reword the question or find another solution if the information seems sensitive, identifying, or personal. AI chatbots are effective tools for increasing productivity. However, you are still in charge of your privacy. In the era of generative AI, a little prudence goes a long way toward protecting your personal and professional life.