Is Claude safe for business data?
Claude Team and Enterprise do not train on your chats by default; on Free, Pro, and Max, Anthropic uses your chats to improve Claude only if you allow it, and keeps them up to 5 years if you do.
Training, plan by plan
- Not used for training: Free, Pro, and Max: not used, unless you allow it.
- Not used for training: Team and Enterprise: not used by default.
What it does with your data
| Question | Free, Pro, and Max | Team and Enterprise |
|---|---|---|
| Trains on your data by default? | Only if you allow it. Anthropic says it uses your chats to improve its models if you choose to allow it, if a chat is flagged for safety review, or if you opt in another way. The setting is under Settings > Privacy, “Help Improve our AI models.” Read on October 23, 2026 · source: privacy.claude.com | No, not by default. Anthropic says it does not train on inputs or outputs from its commercial products. If someone sends thumbs up or down feedback, that conversation may be used. Read on October 23, 2026 · source: privacy.claude.com |
| How long chats are kept | Deleted chats leave your history at once and are deleted from back-end storage within 30 days. If you allow model improvement, Anthropic may keep de-identified data for up to 5 years. Read on October 23, 2026 · source: privacy.claude.com | Chats are kept in the product so people can continue them. Deleted chats leave the history at once and are deleted from back-end storage within 30 days. Chats flagged for a policy violation may be kept up to 2 years. Read on October 23, 2026 · source: privacy.claude.com |
| Who can read them | Not described on the pages we read. Not confirmed · page checked: privacy.claude.com, October 23, 2026 | Your organization’s Primary Owner manages the work account and its data, and can export data that may include members’ conversations and uploaded files. Read on October 23, 2026 · source: privacy.claude.com |
| Where data is stored | Anthropic says your personal data is transferred to its servers in the US, or to other countries outside the EEA and the UK. Read on October 23, 2026 · source: anthropic.com | Not stated for Team or Enterprise on the pages we read. Anthropic points to its Trust Center, which we have not checked. Not confirmed · page checked: privacy.claude.com, October 23, 2026 |
| How to delete | Open the chat, click its name at the top of the screen, and select Delete. To delete several, open Recents, check the chats, and click “Delete Selected.” Read on October 23, 2026 · source: privacy.anthropic.com | Same steps as the personal plans: open the chat, click its name, and select Delete. Deleted chats are removed from back-end storage within 30 days. Read on October 23, 2026 · source: privacy.claude.com |
How to turn training off
- On a personal plan, open Settings > Privacy and check the toggle under “Help Improve our AI models.” Source: privacy.claude.com
- Use Incognito chats for one-off questions. Anthropic says they are not used to improve Claude, even with model improvement on. Source: privacy.claude.com
- On Team or Enterprise, an Owner can turn off the “Rate chats” setting under Organization settings > Data and Privacy, so thumbs up or down feedback is not sent. Source: privacy.claude.com
Safer setup for a small team
Use a Claude Team or Enterprise plan for work. On a personal plan, check the model improvement setting under Settings > Privacy and keep it off for work chats.
Plan to approve: Claude Team or Enterprise
Setting to check: “Help Improve our AI models” (personal plans); “Rate chats” (Team and Enterprise)
A line for your AI policy: Staff may use Claude for work only through our company Team or Enterprise plan. Customer data never goes into a personal Claude account.
Want this in your team’s AI policy? The SafeStart kit, a plain-English AI policy and training for small teams, is coming soon.
Questions about your own setup? Write to hello@terrapilgrim.org.
What changed
No changes recorded since we first published this page on October 23, 2026. Every change goes in the change log.
Sources (7)
- privacy.claude.com/en/articles/10023580-is-my-data-used-for-model-training · read on October 23, 2026
“We will use your chats and coding sessions (including to improve our models) if: You choose to allow us to use your chats and coding sessions to improve Claude”
- privacy.claude.com/en/articles/10023548-how-long-do-you-store-my-data · read on October 23, 2026
“we may retain your data in a de-identified format for up to 5 years in our model training pipelines”
- anthropic.com/legal/privacy · read on October 23, 2026
“your personal data is transferred to our servers in the US, or to other countries outside the European Economic Area”
- privacy.anthropic.com/en/articles/8230524-how-can-i-delete-or-rename-a-convers… · read on October 23, 2026
“Select either "Delete" or "Rename" from the options that appear”
- privacy.claude.com/en/articles/7996868-is-my-data-used-for-model-training · read on October 23, 2026
“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.”
- privacy.claude.com/en/articles/7996866-how-long-do-you-store-my-organizatio… · read on October 23, 2026
“Deleted from our back-end storage systems within 30 days”
- privacy.claude.com/en/articles/9265372-who-owns-and-manages-the-data-of-my-… · read on October 23, 2026
“Your organization's designated Primary Owner manages your Work account and all associated data.”
Related tools
- 3 plans checked
ChatGPT
OpenAI
- Used for training: Free, Plus, and Pro: used, unless you turn it off.
- Not used for training: ChatGPT Business: not used by default.
- Not used for training: ChatGPT Enterprise: not used by default, unless your organization opts in.
Checked October 23, 2026 - 2 plans checked
Gemini
Google
- Used for training: Personal Google Account: used while Keep Activity is on.
- Not used for training: Google Workspace (business accounts): not used without your permission.
Checked October 23, 2026 - 2 plans checked
Microsoft Copilot
Microsoft
- Depends: Copilot app, personal Microsoft account: your choice: you can opt out.
- Not used for training: Microsoft 365 Copilot (work account): not used.
Checked October 23, 2026 - 2 plans checked
Grok
SpaceXSI
- Used for training: Grok app and Grok.com (personal): used, unless you turn it off.
- Not used for training: SpaceXSI API (business and enterprise): not used.
Checked October 23, 2026 - 2 plans checked
Perplexity
Perplexity AI
- Used for training: Free, Pro, and Max: used by default, unless you turn it off.
- Not used for training: Enterprise Pro and Enterprise Max: not used.
Checked October 23, 2026 - 1 plan checked
Meta AI
Meta
- Used for training: Meta AI (personal accounts): used, with no opt-out described.
Checked October 23, 2026
Comparisons
Keep browsing
Get the monthly changes
One email each month: what changed, in plain English.
Never shared. Unsubscribe anytime.