How to stop Claude (Anthropic) from training on your data in 2026

Updated Corrections

At a glance

  • Platform Anthropic Claude
  • Company Anthropic
  • Deletion and retention Deleted chats leave your history immediately and are deleted from Anthropic’s back-end within 30 days, subject to retention exceptions. Memories generated from a chat stay until you delete them in Settings > Memory. Turning off model improvement doesn’t delete saved chats.
  • Official page privacy.claude.com

Claude help articles and Anthropic privacy policies checked October 5, 2026. Anthropic’s Privacy Center articles on training use (updated March 16, 2026) and data storage (updated July 1, 2026) were re-read October 11, 2026; the model-improvement settings article was last read October 1, 2026. Signed-in account interfaces, request submission and outcomes were not tested.

What you need to know first

This guide covers signed-in Claude Free, Pro and Max accounts, including Claude Code used with those plans. Model improvement, Incognito chats, memory, chat deletion, export and account deletion are separate controls. Team and Enterprise plans and the API follow separate documentation.

Choose your task

You don’t need to delete chats or your account to stop future training. Choose what you want to do; each task names the steps below that it needs.

Keep your account but stop future training

Turn off Help Improve our AI models in Settings > Privacy, on the web or in the mobile app. Your saved chats stay, and Anthropic says it stops using your chats and coding sessions in future training runs. Use steps 1 and 2 in the step-by-step list below.

Check the outcome: Reopen Settings > Privacy and confirm the setting is still off. That shows your current setting. It doesn’t show that training already started or completed has changed, and conversations you rate with thumbs up or down, or that are flagged for safety review, can still be used.

Delete stored content

Export first if you want a copy. Then delete chats, and delete memories and projects through their own controls. Use steps 3, 4, 5 and 7 in the step-by-step list below.

Check the outcome: Confirm the chat is gone from your history and the topic no longer appears in Settings > Memory. Anthropic says deleted chats are deleted from its back-end within 30 days, subject to retention exceptions; your account view doesn’t show that step.

Report a Claude response that contains personal data about you

Send a privacy request to privacy@anthropic.com, with or without a Claude account, and include the prompt, response and date. Anthropic may verify your identity, an authorized agent needs proof of authority, and Anthropic says correcting a response may not always be possible.

Check the outcome: Keep Anthropic’s reply and recheck the response later. A reply isn’t proof that a response, a cited web page or a search result has changed, so ask any cited site separately.

Close your account

Only if you want to stop using Claude; it isn’t needed to stop training. Cancel any Pro or Max subscription and wait for its period to end, export your data if you want a copy, then delete the account. Use steps 6, 7 and 8 in the step-by-step list below.

Check the outcome: Keep any confirmation Anthropic shows or sends. Account deletion is permanent, and it doesn’t remove pages or listings about you elsewhere.

Step-by-step: manage your Anthropic Claude data

  1. Step 1: Sign in to your Claude account and open Settings > Privacy.
  2. Step 2: Turn off Help Improve our AI models, then check that the change saved. Your saved chats stay in place.
  3. Step 3: To delete a web conversation, find it in the sidebar or Chats and tasks, open its three-dot menu, choose Delete and confirm. In Chats and tasks, Select lets you delete several at once.
  4. Step 4: Open Settings > Memory to see what Claude remembers. Deleting a chat doesn’t remove memories generated from it, so delete those topics there. To delete every memory, including project memories, turn off Generate memory from chats and choose Reset memory.
  5. Step 5: Delete projects separately if you want to keep the account. Deleting chats doesn’t remove projects or account information.
  6. Step 6: Before deleting a Pro or Max account, cancel the subscription and wait until its current period ends.
  7. Step 7: To keep a copy, choose Export data in Settings > Privacy on the web or Claude Desktop. Exports can’t be started from the mobile apps, and the emailed link expires after 24 hours.
  8. Step 8: To delete the account, open Settings > Account > Delete account and follow the prompts. Account deletion is permanent.

On mobile

  1. In Claude mobile settings, open Privacy and turn off Help Improve our AI models.
  2. For iOS chat deletion, touch and hold a conversation in the chat list, tap Delete and confirm. On Android, open the conversation, use its three-dot menu, tap Delete and confirm.
  3. After any paid subscription has ended, iOS account deletion is Settings > Account > Profile > Delete account. On Android, it’s menu > initials > Profile > Account Actions > Delete Account.

Opt out of AI training

In Settings > Privacy, turn off Help Improve our AI models. Anthropic says this excludes both previously stored and new consumer chats and coding sessions from future training runs. It does not undo training already started or completed. Even after you opt out, Anthropic can use conversations flagged for safety review, and conversations you rate with thumbs up or down, for model improvement. Incognito chats aren’t used for training.

Retention and deletion limits

When model improvement is on, new or resumed chats can remain de-identified in training pipelines for up to five years. Feedback can be kept for five years. For flagged policy violations, inputs and outputs can remain up to two years and classification scores up to seven years. Incognito chats are kept for 30 days by default. Memories aren’t removed when their source chat is deleted. Legal, dispute and policy-enforcement exceptions can apply.

Separate data request

Anthropic’s privacy policies direct rights requests, such as access, deletion or correction, to privacy@anthropic.com, and people without a Claude account can use the same route. Anthropic may verify your identity, and an authorized agent needs proof of authority. If a Claude response contains inaccurate personal data about you, you can ask for a correction; Anthropic says it will make a reasonable effort, but correction may not always be possible. Rights depend on where you live, and Anthropic may decline a request with a lawful reason. Some account deletions need support, as shown in account settings. If several accounts share an email address, say which ones to delete.

Key gotchas

What this doesn't cover

These controls cover your Claude account and Anthropic’s systems. They don’t remove a web page a response cites, a data broker listing or a search result, so ask each source separately. A response about you isn’t proof that your personal data was used in training.

Delist finds exposed personal data, pursues supported removal requests and tracks their status. Claude settings need your own account, and Anthropic may verify your identity for a privacy request. Removing a public listing doesn’t establish that a trained model has changed.

See what personal data is exposed about you beyond Anthropic Claude.

Run a free scan →

Frequently asked questions

Does turning off Claude training delete my chats?

No. Your saved chats stay, and Anthropic says it stops using your chats and coding sessions in future training runs. That doesn’t undo training already started or completed. Delete chats, memories and projects through their own controls.

Does deleting a Claude chat keep it out of training?

Anthropic says a deleted conversation won’t be used for future model training. It doesn’t say deletion changes training already started or models already trained.

Does deleting a chat delete what Claude remembers?

No. Memories generated from a chat stay after you delete it. Delete them under Settings > Memory, or reset memory to delete every memory, including project memories.

Are Incognito chats used for training?

No. Incognito chats aren’t used for training and don’t appear in your history or memory. Anthropic still keeps them for 30 days by default.

Can I ask Anthropic to remove personal data from Claude responses?

You can send a privacy request to privacy@anthropic.com, with or without a Claude account. Depending on where you live, that can include deletion, objection or correction of inaccurate personal data. Anthropic says correcting a response may not always be possible, and it may decline a request with a lawful reason. The request doesn’t remove a website or search result a response cites.

Do these controls remove my personal data from the internet?

No. They cover your Claude account and Anthropic’s systems. Ask each site that shows your personal data to remove it.

The instructions above were checked against the linked documentation, not a signed-in account. Platforms may change their settings layout at any time. This guide is informational and not affiliated with Anthropic.

Check your public exposure

Delist finds exposed personal data, pursues supported removal requests and tracks their status. Claude settings need your own account, and Anthropic may verify your identity for a privacy request. Removing a public listing doesn’t establish that a trained model has changed.

Run a free scan →