← Back to AI & Digital Life

ChatGPT Privacy & Controls

ChatGPT has separate controls for model improvement, Memory, Temporary Chat and connected apps. Understanding the difference makes the settings much easier to use.

Verified September 2026 · Sources

ChatGPT features and settings change. This guide reflects settings verified in September 2026.

The quick version

What you need to know:

"Does ChatGPT use my information?" is more than one question

When you ask this, you might mean: Is it stored? Is it in my history? Can it remember it? Can it train models with it? Can it access my connected apps? The answer depends on which setting you're asking about.

ChatGPT has separate controls for different purposes

Model Improvement: Whether new chats train OpenAI's models. Memory: Whether ChatGPT remembers past context. Temporary Chat: For conversations that don't affect your history. Connected Apps: Access to your email, calendar, or other services.

Turning one setting off doesn't automatically turn off another

Disabling Model Improvement doesn't disable Memory. Disabling Memory doesn't disable app access. Understanding these separate controls gives you more precise choices.

Your account type matters

Personal accounts, workplace accounts, and school accounts have different default settings and protections. Check which account you're using.

Want more help?

Model Improvement (Improve the Model for Everyone)

For personal ChatGPT accounts, OpenAI provides a setting called:

Improve the Model for Everyone

You can manage it in Settings → Data Controls.

What it does

If you turn it off, OpenAI says new conversations can still appear in your chat history but are not used to train ChatGPT.

Important distinction

Chat history and model improvement are separate controls.

What's actually happening

Not universally. How information is used depends on:

  • your account type
  • your settings
  • what feature you're using
  • whether information is manually uploaded or entered
  • whether information comes through certain connected apps
  • whether you submit feedback

For personal accounts, users can turn off model improvement. OpenAI also says Business, Enterprise and Edu workspace data is not used to train generalized models by default.

Memory

Memory is a separate feature from model training.

What Memory does

When enabled, Memory can help ChatGPT use useful context from:

  • chats
  • files
  • connected apps

to personalize future responses. That might include things like:

  • a project you're working on
  • your preferences
  • recurring context
  • information you don't want to explain repeatedly

Memory is managed through Settings → Personalization → Memory.

Memory vs. Training

Turning model improvement off does not automatically mean Memory is off. Turning Memory off does not automatically mean model improvement is off.

Key distinction

Memory and model training are different questions.

Temporary Chat

Temporary Chat gives you another way to control how a particular conversation is handled.

What it does (as of September 2026)

A Temporary Chat:

  • is not used to improve OpenAI's models while it remains temporary
  • does not create new memories
  • can be retained for up to 30 days for safety purposes
  • normally stays out of your chat history unless you choose to save it

Expanded Temporary Chat Controls

OpenAI expanded Temporary Chat controls in August 2026. You may now be able to choose whether a Temporary Chat uses existing personalization such as:

  • Memory
  • custom instructions
  • connected plugins/apps

A personalized Temporary Chat can use existing information without creating new memories.

Important distinction

Temporary does not always mean ChatGPT starts with no information about you. It depends on whether personalization is enabled for that Temporary Chat.

If you save a Temporary Chat

If you choose to save one, OpenAI converts it into a regular conversation. It then follows your normal account settings for:

  • history
  • personalization
  • Memory
  • model-improvement settings

"Temporary" describes how the conversation is handled while it remains temporary.

Connected Apps and Your Data

General Connected Apps

ChatGPT can connect to supported external services. Depending on the feature and account, those may include:

  • email
  • calendars
  • contacts
  • cloud storage
  • workplace or productivity services

Connecting an app lets ChatGPT work with eligible information from that service. That is separate from the question of whether the information trains a generalized model.

Special handling for Google Apps

OpenAI currently gives data directly accessed from connected Google apps special treatment. OpenAI says it does not train generalized models directly on data accessed from connected Google apps, even when Improve the Model for Everyone is on, except in specific situations such as:

  • the user submitting feedback
  • manually copying, pasting or uploading the information
  • certain information becoming part of a generated response

Policies can change, so this should be rechecked periodically.

Core principle

Access and training are separate questions.

Can connected information affect Memory?

Potentially. If Memory is enabled, OpenAI says eligible information from connected services may be used to personalize ChatGPT, subject to settings and feature-specific rules.

Remember:

  • app access
  • Memory
  • model improvement

are different controls.

Work and School Accounts

If you use ChatGPT through work, school, or an organization, your workspace may have different privacy protections or administrator-controlled settings.

Workspace data protection

OpenAI states that Business, Enterprise and Edu data is not used to train its generalized models by default. Organizations may also control:

  • available apps
  • sharing
  • retention
  • permissions
  • workspace features

Always check which account or workspace you are using.

Public Sharing

ChatGPT can create shareable versions of conversations. Before sharing one, review what is included.

What gets shared

A shared conversation link includes everything in that conversation:

  • names
  • personal details
  • uploaded information
  • work information
  • things you forgot were earlier in the conversation

Remember

Private conversation + public share button = potentially public information.

Your Settings Checklist

Depending on your account, you may be able to:

  • turn model improvement on or off
  • enable or disable Memory
  • review your Memory summary
  • use Temporary Chat
  • choose whether Temporary Chat uses existing personalization
  • save or leave a Temporary Chat unsaved
  • connect or disconnect apps
  • delete conversations
  • export data
  • manage sharing

What changes when you turn things off?

Different settings do different things:

Turn Off "Improve the Model for Everyone"

Your regular conversations can remain in history, but new chats are not used for model improvement.

Turn Off Memory

ChatGPT stops using Memory to personalize future conversations according to your current Memory settings.

Use Temporary Chat

The conversation is handled separately from normal chats and is not used to improve models while temporary.

Disconnect an App

ChatGPT loses future access through that connection, though disconnecting an app does not necessarily remove information already present elsewhere.

Quick decision points

MODEL

Do I want new chats used for model improvement?

MEMORY

Do I want ChatGPT to use past context to personalize future responses?

APPS

What outside services have I connected?

TEMPORARY

Would I rather keep this particular conversation separate?

SHARE

Am I creating a link someone else can open?

Put it in context

"Does ChatGPT use my data?" is too broad a question to have one simple answer.

A better question is:

What information is being used, for what purpose, under which setting?

Core principle

Training, storage, history, Memory, connected apps and sharing are different things.

Understanding that gives you much more control than simply choosing between:

  • "AI is private"
  • "Everything I type becomes public."

Neither is a useful description of how the system actually works.

Sources

All sources were accessed and verified during the last review date listed at the top of this guide.