By Craig Reed, Director of Towpath Tech
Artificial intelligence tools can be useful for answering questions, drafting content, researching topics, and simplifying everyday tasks. But using AI also introduces an important cybersecurity consideration: the information you choose to share.
When using an AI tool, treat the information you enter much like information you provide to any other third-party website or service. Different AI services have different privacy practices, and information entered into an AI system may be collected, retained, or processed in ways you may not expect. Artificial intelligence can also create new privacy risks by combining or analyzing information in ways that reveal additional details about an individual.
Keep Personal Information Personal
Before entering information into an AI tool, ask yourself: Would I be comfortable providing this information to an unfamiliar website?
As a general precaution, avoid entering sensitive information such as:
- Social Security numbers
- Bank, investment, or credit card account numbers
- Passwords, PINs, or multifactor authentication codes
- Copies of tax returns or financial statements containing personal information
- Driver’s licenses, passports, or other identification documents
- Confidential medical, financial, or family information
- Personal information belonging to someone else
If you want AI to help explain a document or answer a financial question, consider whether you can remove names, account numbers, addresses, and other identifying information first.
Review Your AI Privacy Settings
Many consumer AI services—including free versions of ChatGPT and Gemini—provide settings that let users opt out of having new conversations used to improve or train AI models. However, these data-sharing or activity settings may be turned on by default, so users should review and change them before entering information. Paying for a consumer plan does not necessarily provide greater privacy by default; the provider, account type, and selected settings determine how information is handled.
While search engines have used search queries and results for decades to improve future searches, AI broadens the scope of information users share. Users can upload legal contracts, financial statements, and other documents for AI to analyze and answer questions about.
Opting out of model training is not the same as deleting your data.
Depending on the service and settings, saved information can still include chat history, prompts and responses, uploaded files, voice transcripts or recordings, images, account and device information, feedback, and limited safety or security logs. Some information may also be retained temporarily after deletion, or longer when required for security or legal reasons.
Saving a chat lets you return to it and pick up where you left off. If you delete it, use a temporary chat, or start a new one, you may need to repeat earlier details. Within the same chat, AI can generally use what you’ve already shared to answer follow-up questions.
Chat history, memory, personalization, and model training have separate settings. Review each one to understand what information is saved and how it is used.
A Good Rule to Remember
AI can be a helpful tool, but it should not automatically be a trusted recipient of your personal information.
Before uploading a document, entering personal information, or acting on something generated by AI, take a moment to consider what you’re sharing, who may have access to it, and whether you can accomplish the same task without providing sensitive information.
