0.01
A Claude account, and your clients' information
Whatever you put into Claude is kept on Anthropic’s servers until you delete it, and your plan decides what they can do with it.
Your chats
Pro or Max
Team
Is it kept?
Yes, until you delete it.
Yes, until you delete it.
Can they train on it?
Yes, unless you turn the setting off.
No, the contract rules it out.
What if you rate a reply?
That chat can be used for training.
That chat can be used for training.
Anything to show a client?
No, it runs on Anthropic's own terms.
Yes, a data processing agreement.
Pro or Max
Is it kept?
Yes, until you delete it.
Can they train on it?
Yes, unless you turn the setting off.
What if you rate a reply?
That chat can be used for training.
Anything to show a client?
No, it runs on Anthropic's own terms.
Team
Is it kept?
Yes, until you delete it.
Can they train on it?
No, the contract rules it out.
What if you rate a reply?
That chat can be used for training.
Anything to show a client?
Yes, a data processing agreement.
Which plan to take depends on how your clients’ information needs to be treated.
Your clients' information
The plan to take
Fine under your own settings
Needs a contract behind it
It needs a contract behind it when a client’s agreement with you asks for one, or when the information is sensitive, such as health or money, and you would want to show where it goes.
On Pro or Max, the training setting is three clicks away:
Step
What to click
1Open your settings
Click your name in the lower left corner of claude.ai or the desktop app, and choose Settings.2Go to Privacy
Choose Privacy from the list, or go straight there at claude.ai/settings/data-privacy-controls.3Turn it off
Under “Help Improve our AI models”, click the switch so it is off.
On the phone app it is the same three steps, your name, then Settings, then Privacy.
It works from then on, so a chat already used for training stays used, which is why it is worth doing before anything about a client goes in.
On either plan, never rate a reply in a chat that holds a client’s details.
A thumb up or down sends that chat to Anthropic as feedback, and they can train on feedback whatever your plan or setting.
ChatGPT works the same way, a personal plan with “Improve the model for everyone” turned off under Settings, then Data controls, or ChatGPT Business, which also starts at two seats.
Keep your login to hand, because Claude Code asks for it the first time you open it.