Check privacy before your first work chat

Allow 5 minutes.

Use invented information for every exercise. Make this check before you type a work prompt.

Personal accounts: Free, Pro and Max

1. Select your name and open Settings.
2. Open Privacy.
3. Find Help Improve our AI models and switch it off for these exercises.

This setting controls model improvement. Anthropic still describes safety-review exceptions. Turning it off does not create a confidential business agreement or remove all storage.

Organisation accounts: Team and Enterprise

Anthropic does not train on commercial inputs and outputs by default. Feedback or an explicit opt-in can be exceptions. Ask your account owner about the organisation's data rules and feedback controls. Do not submit a work conversation with thumbs-up or thumbs-down feedback unless your policy allows it.

A temporary chat is still processed

Incognito chats are not saved to chat history or memory and are not used for training. They are retained for 30 days by default; an organisation's retention setting can mean longer storage. Organisation owners can access them through exports. Incognito is not permission to enter private data.

Try it: five-minute check

Find the Privacy page in your approved account. Record only the setting name and whether it is off. Do not copy account identifiers or take a screenshot of private account details.

Next, write a list of three things you will keep out of practice chats: for example, a client's IRD number, an employee's medical details, and a password.

Warning: A privacy setting does not replace your employer's approval or your obligations under the Privacy Act 2020.

Sources

Settings. Consumer. Business. Incognito.