Claude Privacy Settings: Stop Training, Delete Chats, Export Data & Review Shared Links

Claude gives you separate controls for model training, shared chats, data exports, conversation deletion and Incognito chats. This guide shows what to check, what each setting actually changes, and which privacy controls people commonly confuse.

Everyday AI Guides Last checked: September 3, 2026

Claude privacy settings control several different things: whether eligible consumer chats can be used to improve Claude, which conversations you have shared by link, how to export your account data, and what happens when you delete a conversation. Claude also offers Incognito chats for temporary conversations that stay out of your normal chat history and memory.

The confusing part is that these controls are not interchangeable. Turning off model improvement does not delete your chats. Deleting a chat does not delete your entire account. Incognito does not mean the conversation is retained for zero days. And unsharing a chat does not erase the original conversation from your own history.

Quick answer

For a fast Claude privacy audit, open Claude → Settings → Privacy. Review Help Improve our AI models, check Shared chats for links you no longer want active, and export your data if you want a personal copy. Delete individual chats separately. For a one-off conversation that you do not want added to normal history, memory, or model training, start an Incognito chat.

Important scope This guide focuses mainly on individual Claude Free, Pro, and Max accounts. Claude Team, Enterprise, Claude for Work, and the Anthropic API can have different training, sharing, export, administration, and retention rules.

Claude Privacy Settings: The 5-Minute Audit

Start with the outcome you want rather than the name of a setting. Claude has separate controls for training preferences, stored conversations, shared links, temporary chats, exports, memory, and account deletion.

Your goal Control to use Where to look What it does not do
Stop eligible chats being used in future model-training runs Help Improve our AI models → Off Settings → Privacy Does not delete your chat history or undo training already completed or underway.
Find links you previously shared Shared chats → Manage Settings → Privacy Does not delete the original chat from your account.
Download your Claude information Export data Settings → Privacy on web or desktop Does not move the export into another personal Claude account.
Remove one saved conversation Delete conversation Chat menu or Chats and tasks Does not close your Claude account or automatically manage every other personalization control.
Keep one new conversation temporary Incognito chat Start a new chat outside a project Does not mean Anthropic retains the conversation for zero days.
Control what Claude remembers across chats Memory controls Settings → Memory Memory is a separate control from model-improvement settings.
Permanently close your Claude account Delete account Settings → Account Cannot be treated like a temporary sign-out or reversible cleanup.

A practical order for reviewing your account

  • Decide whether you want your eligible consumer chats used to improve Claude.
  • Review every active shared-chat snapshot and revoke links you no longer need.
  • Export important conversations before deleting anything you may want later.
  • Delete old conversations you no longer want in your saved history.
  • Check Claude Memory separately if your concern is what Claude remembers about you.
  • Use Incognito for future one-off conversations that should stay outside normal history and memory.
  • Only use account deletion if you actually want to permanently close the account.
Privacy control vs. privacy guarantee A setting changes a specific part of how Claude works. It should not be interpreted as a promise that no copy of information can exist anywhere, that every retention exception disappears, or that past model training can be reversed.

How to Stop Claude From Using Chats for Future Model Training

Anthropic gives individual Claude users a setting called Help Improve our AI models. For consumer products such as Claude Free, Pro, and Max, this setting determines whether eligible chats and coding sessions can be used to improve Claude.

On desktop or in your browser

  1. Open Claude and open your account menu.
  2. Select Settings.
  3. Open the Privacy section.
  4. Find Help Improve our AI models.
  5. Turn the setting off if you do not want eligible chats used in future model-training runs.

On mobile

  1. Open your account menu in the Claude mobile app.
  2. Choose Settings.
  3. Open Privacy.
  4. Change the Help Improve our AI models setting.
What turning it off does Anthropic says that after you turn the model-improvement setting off, it will not use new chats and coding sessions for future model training. It also says previously stored chats and coding sessions will stop being used in future training runs.
What turning it off does not do It does not remove information from models that have already been trained, and Anthropic says data can remain involved in training that has already started. Conversations flagged by safety systems may also be handled separately for trust, safety, policy enforcement, or safety research.

This distinction matters because training preference and storage are different questions. If you want a conversation removed from your saved history, delete that conversation separately. If you want to permanently close your Claude account, use the account-deletion controls.

What about Team, Enterprise, Claude for Work, and the API?

Do not apply consumer-plan instructions blindly to organizational or API use. Anthropic states that inputs and outputs from its commercial products are not used for model training by default, although feedback, explicit permission, and other stated cases can be treated differently.

If you use Claude through an employer, school, organization, API account, or managed workspace, check the rules for that specific product and any organization-level policies before assuming your consumer settings apply.

How to Find and Unshare Claude Chats

Claude conversations are private by default, but Claude lets individual users create a shareable snapshot of a conversation. Once you create that share link, anyone with the link can view the shared snapshot.

This is worth auditing because a link you created weeks or months ago can be easy to forget.

Review all of your shared Claude chats

  1. Open Claude → Settings.
  2. Open Privacy.
  3. Find Shared chats.
  4. Select Manage.
  5. Review the title, sharing date, and link for each shared chat.
  6. Select Unshare beside anything that should no longer be accessible through its direct link.
A shared Claude chat is a snapshot Anthropic says the shared snapshot contains messages sent before the chat was shared. Messages added afterward remain private by default. If you later unshare and share the conversation again, the new snapshot can include newer messages.

What happens to attached files?

Anthropic says an attached file itself is not included in the shared snapshot. However, the visible conversation and Claude’s responses are included, so information from a file could still appear if it was discussed or reproduced in the conversation.

Anthropic also says raw data retrieved through MCP tool calls stays hidden from the shared snapshot; the conversation and final outputs remain the parts viewers can see.

Safer sharing habit Before sharing any AI conversation, read the entire snapshot as though you were the recipient. Remove or avoid sharing confidential names, account details, private business information, sensitive personal information, or anything copied from a source you do not have permission to redistribute.

How to Export Your Claude Data

If you want a copy of your Claude information before cleaning up your account, deleting chats, or closing the account entirely, export first.

Anthropic currently makes self-service exports available to individual Free, Pro, and Max users. The export includes conversation data and user data associated with the account.

  1. Open Claude on the web or Claude Desktop.
  2. Open your account menu and select Settings.
  3. Go to Privacy.
  4. Select Export data.
  5. Wait for the download link to arrive at the email address associated with your Claude account.
  6. Sign in to your account and download the export before the link expires.
Available

Web and Claude Desktop

Individual users can start a personal data export from the web app or Claude Desktop.

Not currently available

iOS and Android export request

Anthropic says the self-service export cannot currently be initiated from the Claude iOS or Android apps.

The emailed export link expires Anthropic says the download link expires 24 hours after delivery. If it expires, you can request another export.

An export is useful as an archive, but Anthropic says exported data cannot currently be imported into another personal Claude account or used to migrate conversations between personal accounts.

Team and Enterprise exports work differently

For Team and Enterprise organizations, Anthropic says organization-level data exports are available to the organization’s Primary Owner. Individual organization members do not have the same self-service export option.

How to Delete Claude Conversations

Delete a conversation when your goal is to remove that specific chat from your Claude history. This is different from turning off model improvement, resetting memory, or deleting your whole account.

Delete a Claude chat on the web

  1. Open Claude in your browser.
  2. Find the conversation in the sidebar or open Chats and tasks.
  3. Open the three-dot menu beside the conversation.
  4. Select Delete.
  5. Confirm the deletion.

Claude also supports selecting multiple conversations from Chats and tasks and deleting them together.

Delete chats on iPhone or Android

On iOS, Anthropic says you can touch and hold a conversation in your chat list and choose Delete. On Android, you can delete the open conversation from its menu, and the app also supports selecting multiple chats for deletion.

What happens after you delete a chat? Anthropic says the conversation disappears from your conversation history immediately and is deleted from its backend systems within 30 days.
Deleting one conversation is not account deletion Projects, account information, and other data associated with your Claude account are not automatically deleted merely because you remove one conversation.

If your real concern is what Claude remembers across conversations, review your Memory controls separately. Our dedicated guide explains how to see, edit, delete, pause, or reset Claude Memory.

How Long Does Claude Keep Your Data?

There is no single retention number that accurately describes every kind of Claude data. The answer depends on what the information is, whether you delete it, whether model improvement is enabled, whether you submit feedback, whether safety systems flag the conversation, and which Claude product you use.

Situation Anthropic’s stated treatment Why it matters
You delete a consumer conversation Removed from chat history immediately and deleted from backend storage within 30 days. Deletion is not the same as instant zero-retention.
You allow eligible chats to improve Claude Anthropic says it may retain qualifying data in de-identified form for up to five years in model-training pipelines. Training-related retention can be much longer than ordinary deleted-chat cleanup.
You turn model improvement off Previous and new chats stop being used in future training runs, but training already underway or completed cannot be undone. Preference changes affect future use, not the past.
You use Incognito Not used for training and not saved to normal history or memory, but normally retained for 30 days. Incognito means temporary, not zero-retention.
You submit feedback Anthropic states that data associated with feedback can be retained for five years. Feedback can follow different rules from ordinary conversation use.
A conversation is flagged for certain policy or safety reasons Longer trust-and-safety retention periods can apply. Safety and legal exceptions can override ordinary timelines.
Do not reduce this to “Claude deletes everything after 30 days.” That statement would be misleading. Thirty days applies to important scenarios such as backend deletion of a conversation you delete and default Incognito retention, but Anthropic documents separate retention periods and exceptions for training, feedback, trust and safety, legal requirements, disputes, and other situations.

Anthropic’s current consumer retention documentation also says chats or sessions flagged by automated trust-and-safety systems as violating its Usage Policy can have substantially longer retention periods. This is another reason to treat the privacy controls as specific settings rather than a universal deletion promise.

Claude Incognito Chats: What They Actually Do

Claude Incognito is useful when you want one conversation to stay outside your normal history and memory without changing the settings you use for every other chat.

Anthropic says Incognito chats are available to Claude Free, Pro, Max, Team, and Enterprise users.

Incognito does

Keep the chat out of normal history

The temporary conversation does not appear in your ordinary saved chat history.

Incognito does

Stay out of Claude Memory

Claude does not use its existing memory in the Incognito chat, and the chat is not added to future memory entries.

Incognito does

Stay out of model training

Anthropic says Incognito chats are not used to improve Claude even if Model Improvement is enabled elsewhere in your account.

Incognito does not

Mean immediate deletion

Anthropic says Incognito chats are normally retained for 30 days, with organization-specific retention rules potentially applying on Enterprise plans.

How to start an Incognito chat

  1. Start a new Claude chat outside a project.
  2. Select the ghost-shaped Incognito control.
  3. Confirm that Claude indicates you are in an Incognito chat.
  4. Use the conversation normally.
  5. Save anything important elsewhere before closing it.
You cannot reopen a closed Incognito chat Anthropic says an Incognito conversation cannot be converted into a regular chat or recovered after it is closed.

Incognito still has an important limitation

Anthropic says Claude can still access account-level profile information such as custom styles or personal preferences inside an Incognito chat. So Incognito should not be interpreted as a completely blank or anonymous account session.

Incognito is also currently intended for chats outside projects. If you are trying to manage persistent context rather than a single temporary conversation, use the dedicated Claude Memory controls instead.

How to Delete Your Claude Account

Account deletion is the strongest option in this guide and should only be used when you actually want to permanently close your Claude account.

Export first if you need anything later Anthropic says account deletion is permanent. You lose access to conversations, projects, and other saved account data and cannot recover the account afterward.

Before deleting a paid Claude account

Anthropic currently tells Pro and Max subscribers to cancel the subscription from Billing settings and wait until the current subscription period ends before deleting the account.

Delete your account on the web

  1. Open Claude and select your account name or initials.
  2. Open Settings.
  3. Go to Account.
  4. Select Delete account.
  5. Read the warning and follow the confirmation prompts.

Account deletion is also available through the Claude iOS and Android apps, although the exact menu path differs between platforms.

Before-you-delete checklist

  • Export conversations or account information you may need later.
  • Review your active shared-chat links.
  • Save any project material that does not exist elsewhere.
  • Review Claude Memory if you want to understand what personalization data is currently present.
  • Cancel a paid Pro or Max subscription and follow Anthropic’s current billing instructions.
  • Only confirm deletion once you are comfortable losing access permanently.

Delete, Disable Training, Unshare, Incognito, or Delete the Account?

These Claude privacy controls sound similar because they all affect data in some way. The easiest way to choose correctly is to match the control to the exact outcome you want.

Action Use it when… Chat stays in normal history? Main limitation
Turn off model improvement You do not want eligible chats used in future model-training runs. Yes Does not delete chats or undo training already completed or underway.
Unshare a chat You want to revoke access to a shared snapshot. Yes Does not delete your original conversation.
Delete a chat You want a saved conversation removed. No Does not close the account or automatically handle every other privacy setting.
Incognito You want one new conversation outside normal history, memory, and model training. No Normally retained for 30 days and cannot be reopened after closing.
Manage Memory You want to control persistent information Claude remembers across chats. Separate issue Memory controls are not the same as training or conversation-history controls.
Delete account You want to permanently close your Claude account. You lose access Permanent and not recoverable.

Three Privacy Mistakes to Avoid

Mistake 1

Assuming “training off” means “deleted”

Turning off model improvement changes future training use. It is not a delete button for your conversation history.

Mistake 2

Forgetting old shared links

A chat you shared months ago may still have an active snapshot. Periodically review Shared chats in Privacy settings.

Mistake 3

Treating Incognito as zero-retention

Incognito keeps a chat outside normal history, memory, and training, but Anthropic still documents a default 30-day retention period.

Better habit

Share the minimum necessary data

Privacy settings help, but the strongest everyday control is still avoiding unnecessary secrets, credentials, private records, and unrelated personal data in the first place.

This same principle becomes even more important when an AI can reach outside its own chat window. Before connecting email, cloud storage, calendars, or other services, use our AI Agent Safety Checklist to review access and permissions.

If you let an AI interact with websites or browser tabs, our AI Browser Checklist covers browsing context, form submissions, downloads, and approval boundaries.

Frequently Asked Questions

Does Claude use my chats to train its AI?

For consumer Claude products such as Free, Pro, and Max, Anthropic says eligible chats and coding sessions can be used to improve Claude if you allow model improvement. Safety-flagged conversations and explicit opt-in situations can be treated separately. Anthropic says its commercial products are not used for model training by default.

How do I stop Claude from using my chats for training?

Open Settings → Privacy and turn off Help Improve our AI models. Anthropic says new chats will not be used in future model-training runs, and previously stored chats will stop being used in future runs. This does not reverse models already trained or training already underway.

Are Claude chats private?

Anthropic describes ordinary chats as private by default. If you deliberately create a shared-chat link, anyone who has that link can view the shared snapshot. Organization plans can have different sharing boundaries.

How do I see which Claude chats I have shared?

Individual Free, Pro, and Max users can go to Settings → Privacy → Shared chats → Manage. The shared-chat manager lists previously shared content and lets you unshare snapshots you no longer want accessible.

Does deleting a Claude chat delete it permanently?

Anthropic says deleting a conversation removes it from your visible chat history immediately and deletes it from backend systems within 30 days. Other stated legal, safety, policy, feedback, training, or retention rules can apply to other forms of data.

How long does Claude keep deleted chats?

Anthropic says a conversation you delete is removed from its backend storage systems within 30 days. Do not apply that 30-day number to every category of Claude data because training, feedback, trust-and-safety, organizational, and legal retention rules can differ.

Does Claude Incognito save or train on my chats?

Incognito chats are not saved to your normal chat history, are not added to Claude Memory, and are not used to improve Claude. Anthropic says they are normally retained for 30 days, however, so Incognito should not be interpreted as zero-retention.

Should I export my Claude data before deleting my account?

Yes, if there is anything you may want later. Anthropic says account deletion is permanent. Individual Free, Pro, and Max users can request an export from Claude on the web or Claude Desktop before deleting the account.

Is deleting a Claude chat the same as deleting Claude Memory?

No. Conversation history and Claude Memory are separate controls. If your goal is to review or remove persistent context Claude remembers about you, open the Memory settings and use the appropriate memory control rather than relying only on chat deletion.

Finished the Claude privacy audit? Check what Claude remembers next.

Privacy settings control training preferences, sharing, exports, deletion, and temporary chats. Memory is a separate layer. Continue with the guide that matches what you want to review next.

Keep this guide handy: Claude’s privacy controls and retention rules can change as the product evolves. Recheck the setting you care about before making a consequential privacy or account decision.

Explore more practical guides in AI Safety & Privacy.

This guide is general educational information, not legal advice. Product availability, labels, account paths, retention rules, and organization-level controls can change. Important privacy decisions should be verified against Anthropic’s current official documentation for your specific Claude plan.

Leave a Reply

Your email address will not be published. Required fields are marked *