Keep business data safer when you use Claude

What you will be able to do

You will know what not to paste into Claude, how consumer vs Team/Enterprise training defaults differ at a high level, and where to read Anthropic’s own privacy help—without treating a blog post as a compliance certificate.

The honest answer

Business data in Claude is only as safe as:

• What staff paste or connect
• Which plan and settings you use
• What Anthropic’s current terms say for that product
• How your shop behaves on a busy Friday

No chat tool makes careless pasting safe. Clear rules do.

Training defaults (verify on Anthropic’s pages)

Consumer plans (Free / Pro / Max): Anthropic’s privacy center explains when chats and coding sessions may be used to improve models if you allow model improvement, with separate rules for safety review and feedback. Read the current consumer article: Is my data used for model training?. Incognito chats are described separately in Claude’s help center.

Team / Enterprise (Claude for Work): Anthropic states that for commercial products it does not use your chats or coding sessions to train models by default, unless you opt into programs such as the Development Partner Program. Read the commercial training article and processor/controller help. Your organization may still access workplace conversations under its own admin policies.

Defaults can change. Re-check Anthropic’s privacy center before you set shop policy.

What not to paste

Same postcard rule as customer data and AI tools:

• Card numbers, bank logins, passwords
• Full medical charts or unnecessary health details
• Children’s data, government IDs, unrestricted HR files
• Secrets you would not put in a shared email

Prefer: role + facts needed for the draft; redact names when the job does not need them; use approved connectors with least access — see Drive / Microsoft 365.

Team and Enterprise controls (high level)

Commercial plans add organization-side controls such as admin management of connectors, central billing, and (on higher tiers) deeper retention, audit, and role features as listed on pricing. Owners should decide:

• Which connectors are allowed
• What data classes are forbidden in chat
• Who may share Projects
• How long chats are retained under your agreement

What this article is not

It is not a HIPAA, SOC 2, or legal opinion for your practice. Anthropic may list enterprise offerings (including HIPAA-ready options on some Enterprise paths)—that is their product surface, not automatic coverage for your clinic. Confirm contracts with Anthropic and your counsel.

Try this job: write a one-page Claude data rule

1. List red data (never paste).
2. List yellow data (ok only on Team/Enterprise with admin approval).
3. List green data (approved templates, public pages, anonymized examples).
4. Name who may connect Drive/Microsoft 365.
5. Pin the page in the staff handbook; review quarterly against Anthropic’s current privacy articles.

Short close

Minimize what you paste, match the plan to the sensitivity, and read Anthropic’s current privacy docs—not a remembered rumor.