Keeping clients safe when you use Claude

Your organization's rules come first. If your workplace has a policy about AI tools, follow it. If it doesn't, ask your director before you use Claude for work.

Never put these into Claude

• Names of clients, their family members, or their caseworkers
• Addresses, phone numbers, email addresses
• Birthdays, ages tied to a person, Social Security or ID numbers
• Account, case or file numbers
• Health details, including mental health and substance use
• Immigration status, court cases, or anything about abuse or violence
• Photos or scans of a client's documents
• Anything else that could help someone figure out who a person is

How to ask without private details

1. Describe the kind of problem, not the person. "Someone whose power will be shut off this week" instead of a name and an address.
2. Use a fake name if Claude needs one, like "Client A".
3. Read it again before you press Enter. Could someone who knows your town tell who this is? If so, take more out.
4. Add the real details yourself later, in your own email or case system.

What Anthropic says it does with your chats

Anthropic makes Claude. These quotes are from its Privacy Center, checked on November 3, 2026. Rules can change, so follow the links to read the current pages.

On the free and paid personal plans (Free, Pro, Max):

• Chats are used to improve Claude if "You choose to allow us to use your chats and coding sessions to improve Claude." (source)
• If you allow that, "we may retain your data in a de-identified format for up to 5 years in our model training pipelines." (source)
• When you delete a chat, it is "Removed from your chat history immediately" and "Deleted from our back-end storage systems within 30 days." (source)
• "Your Incognito chats are not used to improve Claude." (source)

On the work plans (Claude for Work, which includes Team and Enterprise):

• "By default, we will not use your inputs or outputs from our commercial products (e.g. Claude for Work, Anthropic API, Claude Gov, etc.) to train our models." (source)

On every plan, the thumbs up and down buttons send the whole chat. "When you provide us feedback via our thumbs up/down button, we will store the entire related conversation ... for up to 5 years." (source)

Even on a work plan, the rule on this sheet stays the same: keep private details out. These policies are about training and storage. They don't make it OK to share a client's information without their consent.

If something private went in by mistake

1. Delete the chat.
2. Tell the person at your organization who handles privacy, the same day. They decide what happens next.
3. Don't click thumbs up or down on that chat.

This sheet is not legal advice. If your organization handles health records, legal cases or other protected information, ask whoever handles compliance before using any AI tool.