How to Turn Off Claude Training on Your Chats in Pakistan
Turn off model improvement in Claude to stop your chats being used for training. See what changes, what stays, how long chats are kept and what Incognito does.
Short answer: Open Claude, go to Settings, then Privacy, and switch off the model improvement setting. After that, Anthropic says it will not use your previous or new chats and coding sessions for future model training. Incognito chats are never used for training. This applies to Claude Free, Pro and Max.
People paste a lot into Claude: CVs, contracts, university essays, business plans. Many do not know there is a setting that decides whether those chats help train the next model. It is worth thirty seconds to check where yours is set.
We read Anthropic's own privacy pages on 12 October 2026. The rules and numbers below come from those pages.
Claude has a model improvement setting for personal accounts
Anthropic's pages call it Model Improvement. It is a switch in your privacy settings. It decides whether Anthropic may use your chats and coding sessions to improve Claude. Anthropic describes it as your choice.
This is about the consumer products: Claude Free, Pro and Max, and Claude Code when it is used on one of those plans. Companies that use Claude for Work or the API are under different terms.
How to turn off model improvement
Open Claude on the web or the mobile app, go to your privacy settings and switch off the model improvement setting. Anthropic links its privacy controls page for this. The same setting exists on desktop and on mobile.
- Sign in to Claude on the web or in the mobile app.
- Open Settings from your profile menu.
- Go to the Privacy section.
- Find the model improvement setting and switch it off.
- Check the switch again to confirm it stayed off.
We could not read the click-by-click steps on Anthropic's page, because that part did not load for us. The wording of the labels may differ slightly on your screen. Anthropic's page points to the privacy controls page in your Claude settings.
What turning it off does and does not change
Turning the setting off stops new and old chats from being used in future training runs. It does not pull your data out of work that has already begun. Anthropic is clear on this point.
| Question | What Anthropic says |
|---|---|
| Are old chats used after I switch off? | Not for future model training |
| Are new chats used? | Not for future model training |
| Models already trained | Your data stays in them, and in training runs already in progress |
| Incognito chats | Never used to improve Claude |
| Chats flagged by safety systems | May still be used for safety work |
The last row matters. Anthropic says that if its safety classifiers flag a conversation, that conversation may still be used to improve internal trust and safety models and to enforce its policies.
How long Anthropic keeps your chats
With model improvement on, Anthropic says it may keep your data in a de-identified form for up to 5 years in its training pipelines. That period applies to new or resumed chats after the setting was turned on. With it off, deleted chats are removed from back-end storage within 30 days.
- Delete a chat and Anthropic says it will not be used to train future models.
- Feedback you send with the thumbs up or down button is kept for 5 years.
- Chats flagged as policy violations can be kept up to 2 years, and classification scores up to 7 years.
The 5 year feedback rule is worth remembering. If you press thumbs down and add your own text, that text can include private details.
Use Incognito for the chats that matter most
Incognito chats are not used to improve Claude, even if model improvement is on. Use them for anything sensitive, like salary details, legal problems or family matters.
The same habit applies to other assistants. If you also use ChatGPT, see how to stop ChatGPT training on your chats. For Google's assistant, see how to delete Gemini chats and turn off activity saving.
At work, the safer rule is to keep client data out of consumer chatbots altogether. The government cyber team's eight rules are in how to use AI chatbots safely at work.
Common questions
Does Claude train on my chats by default?
That depends on your setting. Anthropic describes it as a choice you control in privacy settings, and says Incognito chats are never used. Check your own switch.
Will my old chats still be used if I switch it off?
Not for future training. Anthropic says it stops using your previously stored chats in future model training runs.
Can my data be removed from models already trained?
No. Anthropic says your data remains in model training runs already in progress and in models already trained.
Does this cover Claude for Work or the API?
No. Anthropic's pages cover Claude Free, Pro and Max. Business and API products have separate terms.
How long are deleted chats kept?
Anthropic says deleted conversations are removed from its back-end storage within 30 days.
Is Claude free to use in Pakistan?
Claude has a free plan. What it includes is compared in the free plans of ChatGPT, Gemini and Claude in Pakistan.
How we verified this
What we checked, where we read it, and what we could not confirm.
We read Anthropic's privacy pages Is my data used for model training?, How long do you store my data? and How do I change my model improvement privacy settings? on 12 October 2026. We could not read the step by step screens on the last page. Anthropic may change these terms, so check the pages again.
About the author
Technology Journalist & Multimedia Producer
Ahmad Ali is a technology journalist and multimedia producer at Pakistan Era, joining the team in October 2026. He covers consumer technology and digital life for Pakistani readers, combining research with hands-on product testing.
