What Should I Share With AI?

NEED → MINIMIZE → CHECK SETTINGS

This is how you share intentionally. You're giving the AI what it needs to help you, while keeping unnecessary information private.

Information Sensitivity Levels

Lower Concern: Public information, general questions, brainstorming, basic information already shared publicly. Example: "How do I make pasta?" or general information about a public figure.

Think First: Personal preferences, opinions, descriptions of situations where privacy matters. Examples: Health concerns, work situations, family dynamics. These are reasonable to share depending on context and tool, but consider: Do I want this used to train the model? Will I regret sharing this later?

Higher Sensitivity: Passwords, Social Security numbers, account information, financial details, medical information, information about other people without consent, proprietary work information. These generally should not be shared with AI tools unless absolutely necessary, and you've considered the risks.

Use this framework with each tool. Different tools have different protections and default settings.

Training is one privacy question. But there are others:

• Is the conversation retained or deleted?
• Can humans review what I shared?
• Are connected apps accessing other information?
• Can I access or delete my data later?
• Where is my data stored?

Different tools handle these differently. Check the tool's privacy settings and documentation.

Connected apps can share more than you realize

Many AI tools can connect to other apps and services—email, calendar, cloud storage, browsing history, photos.

Ask yourself: What can it access? What will it do with that access? Can I disconnect it later?

You can usually turn off connected-app permissions in the tool's settings.

Public sharing is different from private

If you use a "share link" or "public share" button in an AI tool, your conversation (or part of it) becomes public.

Remember: Private conversation + public share button = potentially public information.

ChatGPT Settings (OpenAI)

ChatGPT has a "Data Controls" setting in Settings.

• "Improve the model for everyone": When ON, chats may train models. Turn it OFF to prevent this (personal-user setting).
• Memory: ChatGPT remembers details across conversations. Separate from model training. You can disable it.
• Temporary Chat: Retained 30 days then auto-deleted. Not used for training while temporary. Can still use existing personalization.

Google Gemini Settings

Gemini has a "Keep Activity" setting.

• Keep Activity ON (default): Chats saved and may improve Google services and AI models.
• Keep Activity OFF: Future chats not used for training (except feedback). Chats won't appear in activity; features requiring saved activity may be unavailable.

General Strategies

• Don't paste full passwords or account numbers. Share only what's needed or redact sensitive parts.
• Use pseudonyms instead of real names when describing situations.
• Review connected-app permissions regularly. Disconnect any you're not using.
• If sharing higher-sensitivity information, consider whether a public AI tool is right, or if the tool's privacy protections are sufficient.

Put it in context

Share intentionally. Give AI enough context to help without automatically giving it everything.

You wouldn't tell a stranger your password. You wouldn't email medical history to a random business. Apply the same principle to AI tools:

Do I have the right to share this with this tool? Is this information mine to share? Different answers lead to different choices. That's how you stay in control.