Using AI chatbots safely: privacy, what not to paste in and your data settings

TUREFF• •0

AI chatbots are easy to talk to, and that is exactly why people tell them far more than they would type into a search box. A conversation feels private, but it is text sent to a company's servers, stored for some period and, depending on the service and your settings, possibly read by staff or used to improve future systems. A few habits let you get the benefit without handing over things you will regret.

What happens to what you type

The details differ between services and change over time, but the general pattern is similar:

Your conversations are stored, at least for a while, so that your history appears when you come back.
They may be used to train or improve the system unless you switch that off. Many consumer services offer a setting for this; business and education accounts often have different defaults set by the organisation.
Some conversations may be reviewed by people, for example to check for abuse or to assess quality.
Deleting a chat removes it from your view, but the provider may keep copies for a period for safety or legal reasons. The privacy policy usually says how long.

Read the privacy page of the service you actually use rather than relying on what a friend says about another one.

What not to paste in

A good rule is to treat a chatbot like a stranger's notebook: if you would not want it read aloud, leave it out.

Passwords, PINs, recovery codes and security answers. There is never a reason for a chatbot to know these.
Identity and financial numbers: passport, national ID or resident registration numbers, bank details, card numbers.
Other people's personal information: a colleague's medical situation, a friend's messages, a customer list. They have not agreed to share it.
Confidential work material: unreleased plans, contracts, source code your employer has not cleared, client data. Many workplaces have explicit rules on this.
Full medical or legal documents with your name and details on them. If you want help understanding one, remove the identifying parts first.

How to ask without oversharing

You can usually get the same help with less detail.

Replace names with roles: write my manager or Client A instead of real names.
Change or remove numbers that are not needed for the question.
Summarise rather than paste: describe the situation in your own words instead of uploading the whole email chain.
Ask the general question first. How do I respond to a late payment reminder? often works without the reminder itself.

Settings worth checking

Look through the settings menu of whichever chatbot you use. Common options include:

A switch for whether your chats may be used to improve the model. Turn it off if you would rather they were not.
Temporary or incognito chats that are not saved to your history and, on some services, are not used for training.
Memory or personalisation features that let the assistant remember facts about you across conversations. Check what it has stored and delete anything you do not want kept.
History deletion and data export, so you can see and remove what is held.
Connected apps and file access. If you have linked email, cloud storage or a calendar, review what the assistant can reach.

Names and locations of these settings change regularly, so if you cannot find one, search the provider's help pages.

Account and device safety

Use a strong, unique password and two-step verification on your chatbot account, because your history may be revealing.
Sign out on shared computers. Anyone who opens the browser may see your past chats.
Be wary of copycat apps and browser extensions claiming to be a well-known chatbot. Download from the official website or app store listing.
Be careful with shared links. Sharing a conversation can make it viewable by anyone with the link.

Watch out for scams using AI

Fraudsters use chatbots too, to write convincing messages in fluent language. Polished grammar is no longer a sign that an email is genuine. Verify unexpected requests for money or codes through a channel you already trust. Have you changed any privacy settings on the chatbots you use? Share your tips in the comments.