AI Privacy: What to Check Before You Paste Data

The quickest way to leak something you shouldn’t is a copy-and-paste into a chatbot at the end of a long day. AI tools are so easy to use that it’s easy to forget the text you paste leaves your computer and lands on someone else’s servers.

You don’t have to stop using AI to stay safe. You need a short routine: know your tool’s settings, know what never to paste, and know how to strip out the sensitive parts when you do need help with a real document.

Step 1: Know which plan you’re on

The same AI product can handle your data very differently depending on the plan.

• Free consumer plans often have the loosest terms. Some use conversations to improve their models unless you turn that off.
• Paid personal plans may offer more control, but the defaults vary by provider.
• Business and enterprise plans usually come with contractual promises: your data isn’t used for training, admins control retention, and there may be compliance certifications.

If your employer provides an AI tool, use that one for work. A personal free account, even if it’s the same brand, may not carry the same protections.

Step 2: Check four settings

Open your account’s privacy or data settings and look for these. The menu names change often, so search the help pages for each term if you can’t find it.

1. Training on your data. Can you opt out of your chats being used to improve the model?
2. Chat history and retention. How long are conversations stored? Can you delete them, or use a temporary chat that isn’t saved?
3. Memory or personalization. Some assistants remember facts about you across chats. Check what’s stored, and clear anything you don’t want kept.
4. Connected apps. If you’ve linked email, cloud storage or calendars, review what the AI can access and remove anything you don’t use.

Do this once, then recheck every few months. Providers update policies, and new features sometimes arrive switched on.

Step 3: Never paste these

Some things shouldn’t go into any AI tool, whatever the settings say:

• Passwords, API keys, access tokens and recovery codes
• Government ID numbers, passport details and tax numbers
• Full card numbers, bank account numbers and PINs
• Other people’s medical or financial records
• Anything covered by a confidentiality agreement or your employer’s AI policy
• Unreleased financial results or deal details, if you work at a listed company

If a task seems to need one of these, it almost never does. The AI needs the shape of the information, not the real values.

Step 4: Anonymize before you paste

When you need help with a real document, such as a contract, a complaint email or a customer list, swap the sensitive parts for placeholders first.

Replace this	With this
Names of people	[PERSON A], [PERSON B]
Company names	[CLIENT], [SUPPLIER]
Emails and phone numbers	[EMAIL], [PHONE]
Exact amounts (if sensitive)	[AMOUNT] or a rounded figure
Addresses	[CITY] or remove entirely
Account, order or case numbers	[REF-1]

Watch for details that identify someone in combination. “The only female engineer in our Lahore office” points to one person even without a name.

You can ask the AI to help write the anonymized version, but only after you’ve removed the obvious identifiers yourself. A useful follow-up prompt:

Here is a document where I've replaced names and contact details with
placeholders. Before we start, point out any remaining details that could
still identify a person or company, so I can remove them.

Step 5: Think about what comes back

Privacy isn’t only about what you send. Check the output too:

• Don’t paste AI output straight into public places without reading it. It can repeat details from your input in ways you didn’t expect.
• Treat generated code with care if it handles personal data. Review how it stores and logs information.
• Be careful with shared chats. Many tools let you share a conversation by link. Anyone with the link can usually see everything in it, including what you pasted.

A two-minute checklist

Before pasting anything that isn’t public, ask yourself:

1. Am I using the right account (work tool for work data)?
2. Is training on my data switched off, or covered by a business agreement?
3. Have I removed passwords, ID numbers and payment details?
4. Have I replaced names and contact details with placeholders?
5. Would I be comfortable if this text were read by my manager or client?

If any answer is “no,” fix that first. It takes less time than explaining a data leak.

When to skip the cloud entirely

For truly sensitive material, such as legal files or patient records, the safest option may be an AI model that runs on your own computer, where nothing leaves your machine. Small open-weight models have become capable enough for summarizing and drafting. See our guide on how to run an AI model on your laptop, and our comparison of open-weight vs closed models for the trade-offs.

Key takeaways

• The plan you use matters: business plans usually offer stronger data protections than free ones.
• Check four settings: training, history, memory and connected apps.
• Never paste passwords, ID numbers, payment details or other people’s private records.
• Anonymize real documents with placeholders before asking for help.
• For the most sensitive work, consider a local model that keeps everything on your device.

Working with long documents once they’re cleaned up? Our guide to summarizing long documents with AI shows how to get accurate results.