Millions of people use AI tools daily without thinking much about what happens to their conversations. The honest answer isn't terrifying — but it's worth understanding before you share anything sensitive.
This guide explains what each major AI platform actually does with your data, in plain English, so you can make informed choices about what to share and what to keep offline.
What AI Apps Actually Collect
Every major AI app collects some version of the following:
- Your conversations — everything you type and the responses you receive
- Account information — name, email, payment details if you pay
- Usage data — how often you use it, what features you use, device type
- Feedback you provide — thumbs up/down on responses
This data is typically used for: providing the service to you, safety monitoring (detecting harmful use), and training future AI models (though opt-out options exist).
The risk isn't that your conversations are being sold to advertisers — the major AI companies don't do this. The risk is that conversations could be accessed by company employees for review, or that data breaches could expose stored conversations.
ChatGPT (OpenAI) Privacy
What OpenAI collects and does with it
- Stores your conversations on their servers
- By default, uses conversations to train future models (can be turned off)
- Employees can review conversations for safety purposes
- Conversations persist until you delete them (or use Temporary Chat)
Privacy settings to enable: Go to Settings → Data Controls → turn off "Improve the model for everyone." This opts you out of your conversations being used for training. Use Temporary Chat (toggle in the top bar) for sensitive conversations — these aren't saved to your history.
OpenAI's full privacy policy: openai.com/policies/privacy-policy
Claude (Anthropic) Privacy
What Anthropic collects
- Stores conversations; doesn't use free-tier conversations for training by default
- Maintains conversations for 30 days after you delete them (for safety purposes)
- Strong constitutional AI approach — less likely to be used for advertising data
- Anthropic has a more privacy-forward culture than some competitors
Privacy settings: Claude's privacy controls are available in account settings. Review Anthropic's privacy policy at anthropic.com/privacy for current retention details — these policies update periodically.
Google Gemini Privacy
What Google collects
- Conversations are associated with your Google account and subject to Google's broader privacy framework
- Google reviewers may read conversations to improve AI (human review program)
- Conversations may be retained for up to 3 years by default
- If you use Gemini in Google Workspace, data governance depends on your organization's settings
Privacy settings: In Gemini settings, turn off "Gemini Apps Activity" to stop conversations being saved to your Google account. Note that turning this off may reduce personalization. Review Google's Gemini privacy FAQ for current settings.
Microsoft Copilot Privacy
What Microsoft collects
- Consumer Copilot conversations are stored and may be reviewed for safety
- Microsoft 365 Copilot (enterprise) has stronger contractual data protection
- In Bing, conversations are associated with your Microsoft account if signed in
- Privacy controls available in Microsoft account settings
For personal use, Copilot's privacy is comparable to ChatGPT. For workplace use, the M365 enterprise version offers better data protection guarantees.
Simple Rules to Stay Private
Never share with AI:
- Passwords or security codes
- Bank account, credit card, or SIN numbers
- Medical records or highly sensitive health information
- Legal documents under active litigation
- Confidential client information (check your company's AI policy)
- Information about people who haven't consented to being discussed
Safe to share: general descriptions of your situation without identifying details. Instead of "My employee John Smith at [company] is having performance issues," say "I need to write a performance improvement plan for an employee who has been missing deadlines." You get the same useful output without the identifiable details.
Privacy Settings to Enable Now
Take 5 minutes to do this for any AI app you use regularly:
- ChatGPT: Settings → Data Controls → Disable "Improve the model for everyone"
- Gemini: My Activity → Turn off "Gemini Apps Activity"
- Claude: Review account settings and current privacy policy
- Microsoft Copilot: Microsoft account → Privacy dashboard → Review AI activity settings
These settings don't make AI conversations completely private, but they meaningfully reduce data retention and training data use.
For more on using AI safely, see 30 ways to use AI in daily life. For tool comparisons, best AI apps for personal use covers the major options with honest assessments.
Frequently Asked Questions
Related Resources
- What is an AI agent? Plain-English guide Understand the tech behind the tools you use.
- Find the right AI for your lifestyle Match an AI agent to how you actually live.
- Tech guides for non-technical people Step-by-step help getting started with new tools.
Can AI companies see my conversations?
Yes, in most cases. Employees at OpenAI, Anthropic, and Google can potentially access conversations for safety review. Opt-out settings reduce training data use but don't eliminate the possibility of review for safety purposes. For highly sensitive information, don't share it with consumer AI apps regardless of privacy settings.
Is ChatGPT safe for personal use?
Safe for everyday tasks: writing, research, planning, learning. Avoid sharing passwords, financial account credentials, government ID numbers, medical records, or confidential work information. Enable the data controls in settings to opt out of training data use.
Which AI has the best privacy?
Anthropic (Claude) has strong privacy commitments and doesn't use free-tier conversations for training by default. For maximum privacy, use ChatGPT's Temporary Chat mode or Claude with privacy settings enabled. Enterprise versions of all platforms offer stronger contractual data protection.
Should I use AI for work conversations?
With caution. Check your organization's AI policy before sharing anything work-related. Avoid confidential client information, proprietary strategies, or anything under NDA. For workplace use, enterprise AI tiers (ChatGPT Enterprise, Claude for Work) offer stronger data protection than consumer apps.
Does AI store my conversations forever?
Default retention varies by platform and changes with policy updates. ChatGPT stores conversations until you delete them (or use Temporary Chat). Most platforms let you delete your history. Check each platform's current settings — policies update regularly, so verify against the current privacy policy rather than outdated articles.