Think of a chat as a postcard, not a diary
A helpful rule of thumb is to treat what you type into an AI assistant as something that might be stored, read by others under certain conditions, or kept longer than you expect. That is not a claim about any particular product. Practices differ widely between services, and they change. But it is a safer starting point than assuming a chat window is a private conversation with a trusted friend.
Typed text, uploaded files, images, and voice recordings may all be handled by the service. Depending on the product and its settings, that material might be kept in your account history, reviewed by staff or contractors for safety or quality purposes, or used to improve the system. Some services let you opt out of some of these uses, and some offer business plans with different terms.
What to avoid typing in
When in doubt, leave it out. Be especially careful with:
- Passwords, passcodes and recovery codes. Never share these with an assistant. A password manager is the right place for them.
- Financial details. Card numbers, bank account numbers, tax identifiers and similar information.
- Government identification numbers. Things like passport, driver’s license or social-security-type numbers.
- Health information. Diagnoses, test results, prescriptions or records tied to you or someone else. It is generally better to ask in general terms.
- Private details about other people. Friends, relatives, coworkers and especially children have not agreed to have their information shared with a service.
- Confidential work material. Client data, unpublished plans, source code, contracts, internal documents. Many employers have rules about this, so check yours before using any outside tool.
- Precise location and routines. Your home address, travel dates and daily schedule.
- Anything you would not want to see leaked. Data breaches happen to organizations of all sizes, and stored chats are not immune.
If you need help with a document that contains sensitive details, consider removing or replacing names, numbers and addresses with placeholders first. For example, write “[Client A]” or “[account number]” in place of the real details.
Check your data settings
Look for these controls in the account settings or privacy section of any assistant you use. Names and options differ, so treat the list as things to hunt for.
- Chat history. Can you turn it off, view it, or delete past conversations? Is deletion immediate or delayed?
- Use of your content for improving the service. Is there a switch to opt out of your conversations being used for training or product improvement?
- Temporary or incognito modes. Some tools offer a mode that does not save conversations to your history. Read what it promises and what it does not, since the service may still keep data for a limited time.
- Voice recordings. For voice assistants, can you review and delete audio, and stop it from being retained?
- Connected accounts and permissions. Does the assistant have access to your email, calendar, files, contacts or location? Grant only what you need.
- Memory or personalization features. Some assistants remember details across conversations. Find out what is remembered and how to edit or clear it.
- Sharing links. If you can share a conversation through a link, know who can open it and how to revoke it.
- Data export and deletion. Look for ways to download or erase your data and for what happens when you close the account.
Then read the privacy policy, at least the sections about what is collected, how long it is kept, who it is shared with, and how to make requests about your data. Policies are long, so searching for words such as “retention,” “training,” “share” and “delete” can speed things up.
Account security counts too
Your conversations are only as private as your account. Use a strong, unique password, ideally stored in a password manager, and turn on two-factor authentication where it is offered. Be careful with shared computers, and sign out when you are finished. Browser extensions and apps that claim to add AI features can request wide access to what you do online, so install only what you trust and review their permissions. Fake assistant apps and websites also exist, and phishing messages sometimes imitate popular tools to collect logins.
Things to keep in mind about work and school
If you are using an assistant for work, an organization-provided account may come with different data handling than a personal one. Using a personal account for confidential material can break workplace policies. Schools and universities also set their own rules on AI use and on what students may upload. When rules are unclear, ask.
Privacy and the people around you
Photos, voice recordings and messages often include other people. Before uploading a group photo, a recording of a meeting or a screenshot of a chat, consider whether everyone in it would be comfortable with that. In some places, recording conversations without consent is regulated by law. Those rules vary, so consider checking the ones that apply where you live.
For parents, it helps to talk with children about not typing real names, schools, addresses or photos into chatbots, and to look at the settings of any assistant they use. Parental controls vary in what they can cover.
Where the data goes after it leaves your device
Where processing happens, on the device or in the cloud, influences exposure. Features that run locally may not need to send your input anywhere, while cloud features do. The difference is explained in the article on on-device AI. Even so, assume that anything sent to a server could be stored for some time.
A simple routine
- Decide before you type: would I be comfortable if this were stored or read by someone else?
- Remove names and numbers you do not need to share.
- Set history and training options to match your comfort level.
- Review connected apps and permissions every so often.
- Delete old conversations you no longer need.
- Protect your account with a strong password and a second step.
No setting can guarantee privacy, and policies can change, so the safest information is the information you never entered. Within that limit, a few careful habits let you use AI assistants for everyday tasks while keeping risks reasonable. If you have specific legal or regulatory concerns, such as handling other people’s personal data in a business, a qualified professional can advise on your situation.