You can stop all three AI tools from training on your chats, and it takes less than five minutes per app. Each one buries the setting inside a menu, and all three default to on for personal accounts the moment you sign up.
This guide walks you through the exact steps for ChatGPT, Claude, and Gemini. I will also show you what these settings actually do, what they cannot undo, what the companies do not advertise clearly, and what you should never type into any AI chatbot, regardless of your privacy settings.
This is for anyone using ChatGPT Free, Plus, or Pro; Claude Free, Pro, or Max; or Google Gemini on a personal account. If you are on a business or enterprise plan, your data is usually protected by default, and I will confirm that for each tool below. If you are somewhere in between, keep reading because the details matter.
Why the Default Is the Problem
Most people assume their chats are private. They are not, unless you go and find the toggle and flip it yourself.
Think about what you have already typed. Job applications, health questions, salary negotiations, business ideas, relationship advice. Every bit of that, on a personal account, with default settings, is potentially training data.
The companies are not hiding this. It is in their terms. But those terms run long, the privacy settings are buried two or three clicks deep, and no pop-up on the home screen reminds you to check.
That is the design problem. And it is your job to fix it now, for your own account.
There is also something important to understand before you start. You cannot remove data from AI models that have already been trained on it. Once it is baked into a model’s weights, there is no delete button for that. According to information from OpenAI’s privacy documentation, the training opt-out is forward-looking. What you can do today is stop the companies from using your future chats. The faster you act, the less data goes in.
How to Stop ChatGPT From Training on Your Data
Does ChatGPT Use Your Data by Default?
Yes. On Free, Plus, and Pro accounts, OpenAI uses your conversations to improve future models by default. That setting is on when you arrive, and you have to go turn it off yourself.
ChatGPT Team, Enterprise, and Edu accounts are excluded from training by default. If you are on one of those, you can skip this section. API users are also excluded automatically under OpenAI’s zero-retention policy, which means API inputs and outputs are not stored or used for training at all.
Does ChatGPT Plus Use Your Data for Training?
Yes. Paying for Plus does not automatically give you more privacy. The training toggle is still on. The subscription buys you faster speeds, priority access to new models, and higher usage limits. It does not change how your data is handled. You still have to go find the setting and switch it off yourself.
Step-by-Step: Turn Off ChatGPT Model Training
This takes about 60 seconds.
Step 1. Log in at chatgpt.com. If you use ChatGPT without logging in, data is collected for training regardless of any account settings.
Step 2. Click your profile icon. Depending on your version, this sits in the top-right or bottom-left corner of the screen.
Step 3. Select Settings.
Step 4. Go to Data Controls.
Step 5. Find the toggle labeled “Improve the model for everyone” and switch it off.
Step 6. While you are in Data Controls, look for the voice and video training toggles and turn those off too. They are separate from the main one.
Once you do this, your new conversations are excluded from future training runs. The setting applies to your entire account across every device, so you do not need to repeat it on your phone or in the desktop app.
Use OpenAI’s Privacy Portal as a Second Layer
There is an additional step most guides skip. OpenAI also offers a route through its Privacy Portal where you can select “Do not train on my content.” You can reach this by going to Settings, then Data Controls, then clicking “Learn more.” Doing both gives you the strongest protection available on a personal account.
Use Temporary Chat for Sensitive Conversations
Even with training switched off, the best habit for anything sensitive is to use Temporary Chat. Temporary chats do not appear in your history, are not used for training, and OpenAI keeps them for no more than 30 days for safety reasons.
To start one, open the model selector at the top left of the chat interface and toggle Temporary Chat on. Use it whenever you are typing something you would not want stored.
What This Setting Does Not Do
Turning off “Improve the model for everyone” stops future training. It does not delete your chat history. Your conversations stay visible in your sidebar exactly as before. The setting does not undo anything already used in a completed training run. And it does not cover a browser extension, a third-party integration, or a colleague who pastes your document into their own account. The opt-out is scoped to your account, not to your data everywhere it might travel.
How to Stop Claude From Training on Your Data
What Changed in August 2025
This is the section that surprises people most, especially longtime Claude users.
For two years, Anthropic held a position that set Claude apart. While other companies trained on user data by default, Anthropic did not. Your chats were deleted after 30 days. Nothing was used for training unless you submitted explicit feedback. That privacy-first approach was a real differentiator, and many people chose Claude specifically because of it.
That changed in August 2025.
According to information from Anthropic’s official data policy documentation, starting August 28, 2025, Anthropic began giving Free, Pro, and Max users the choice to allow their data to be used to improve future Claude models. This was described as a choice, but the revised terms meant that if you accepted the update without touching the setting, you were opted in. The deadline to make a selection was September 28, 2025, and new users now see the choice during sign-up.
Claude.ai moved from an opt-in to an opt-out consent model. Previously, your data was only used for training if you explicitly allowed it. Now it is used unless you explicitly disable it.
If you have been using Claude since that update and never checked your privacy settings, your chats have likely been included in training data.
Does Claude Train on Your Data by Default?
On consumer plans, meaning claude.ai Free, Pro, and Max accounts, Anthropic does train on your chats and coding sessions by default since the August 2025 terms change, unless you opt out.
On commercial products, including the API, Claude for Work, Claude Enterprise, Education, and Government plans, your data is not used to train models. That protection is built into the contract.
One more thing worth knowing: paying for Claude Pro or Max does not change how your data is handled. The plan price buys usage limits and feature access, not different privacy terms. Both Max and Pro sit under the same consumer terms as the free plan.
How Long Does Claude Keep Your Data?
This is where the numbers get significant. If the training toggle is on, Anthropic may retain your data in a de-identified format for up to five years. If you opt out, retention drops to approximately 30 days. Deleted conversations are not used for training under any circumstance, according to Anthropic’s retention documentation, but deleting a conversation from the UI and turning off training are two separate actions.
Step-by-Step: Turn Off Claude Model Training
This also takes about 60 seconds.
Step 1. Go to claude.ai and sign in.
Step 2. Click your name or profile icon in the bottom-left corner.
Step 3. Select Settings.
Step 4. Click the Privacy tab.
Step 5. Find “Help improve Claude” (also labeled “Help Improve our AI models” in some versions) and toggle it off.
You can also go directly to claude.ai/settings/data-privacy-controls if you want to skip the navigation.
Opting out stops your new conversations from being used for training and also stops previously stored conversations from going into future training runs. It is not retroactive for training that has already been completed or is already in progress, but it covers everything from that point forward.
Does Claude Use Your Coding Sessions Too?
Yes. Both regular chat conversations and Claude Code sessions fall under the same training policy on consumer plans. If you use Claude for programming help, build side projects with it, or debug code, those sessions are included unless you opt out. The privacy toggle covers both.
Use Claude’s Incognito Chat for the Most Sensitive Topics
Incognito chats are never used for model training regardless of how your main toggle is set. Anthropic still keeps them for approximately 30 days for safety purposes, and conversations flagged for safety review can still be used to improve safety classifiers. But for day-to-day privacy, Incognito mode is the safest option on Claude.
To start an Incognito chat, click the new chat icon and look for the Incognito option at the top of the interface.
Are Claude Chats Confidential?
Partially. According to Anthropic’s published documentation, they use a combination of automated tools and processes to filter sensitive data, and they state that no information is shared with third parties. However, that does not mean your chats are invisible to Anthropic internally. Conversations flagged for potential policy violations may be reviewed by human moderators and retained longer for safety purposes, regardless of your training toggle setting.
How to Stop Gemini From Training on Your Data
How Gemini Handles Your Data
Gemini is different from ChatGPT and Claude in one important way. It is built directly into your Google account, which means the privacy setting is not inside the Gemini app itself. You manage it through your Google account settings.
The setting is called “Gemini Apps Activity,” which Google is in the process of renaming to “Keep Activity” in some regions. The two names refer to the same thing. The process is the same regardless of what it is called in your account.
When this setting is on, your conversations are stored in your activity history, may be used to improve Google’s AI models, and a sample of them may be read and rated by human reviewers. This applies not just to text, but also to files, photos, videos, and screen captures you share in Gemini, according to Google’s own product communications.
Chats that have entered human review are kept by Google for up to three years, even if you later delete your activity. That is the detail that changes how urgently you should act. Simply deleting your chat history after the fact is not the same as having turned the setting off before your chats reached a reviewer.
Step-by-Step: Turn Off Gemini Apps Activity (Desktop)
Step 1. Go to myaccount.google.com.
Step 2. Click “Data and Privacy” in the left sidebar.
Step 3. Scroll to “History settings.”
Step 4. Find “Gemini Apps Activity” or “Keep Activity” and click it.
Step 5. Select “Turn off.”
You can also reach this directly by going to myactivity.google.com/product/gemini. From there, click the Keep Activity card at the top and switch it off.
Step-by-Step: Turn Off Gemini Apps Activity (Mobile)
Step 1. Open the Gemini app.
Step 2. Tap your profile picture in the top-right corner.
Step 3. Tap Gemini Apps Activity.
Step 4. Select Turn off and confirm.
You only need to do this once. The change applies across all devices where you are signed into that Google account.
What Happens When You Turn Off Gemini Chat History?
This is where Gemini works differently from ChatGPT and Claude. On ChatGPT, you can turn off training and still keep your full chat history in the sidebar. On Gemini, training and history are linked inside the same activity setting. When you turn off Keep Activity, your new chats are no longer saved to your conversation list either.
If you want records of specific conversations, copy and save them manually before turning the setting off.
Google still keeps new chats for up to 72 hours after you turn the setting off. This is for service provision, abuse monitoring, and safety processing, not for model training.
Delete Your Existing Gemini Activity Too
Turning off the toggle stops future data collection. But your older chats are still stored. On the same activity page, you can delete past entries individually, by date range, or all at once. You can also set up auto-delete for future data. Keep in mind that conversations already seen by human reviewers are stored separately for up to three years and may not be fully covered by this deletion.
Use Gemini Temporary Chat for Sensitive Topics
Gemini also offers a Temporary Chat mode. These conversations do not appear in your activity history, do not personalise your Google account, and are not used for model training. Google keeps them for up to 72 hours for processing purposes and then deletes them. This is the safest mode for sensitive queries in Gemini.
What About Gemini Inside Gmail and Google Docs?
If you use Gemini features inside Gmail, Google Docs, Drive, or other Workspace apps, check your Smart Features settings separately. Go to Gmail Settings, click “See all settings,” and look under the General tab for Smart Features and Personalization toggles. These are separate from the Gemini Apps Activity setting. According to information from Google’s support documentation, Workspace content including Gmail and Docs is contractually excluded from training Google’s foundation models on managed accounts, but personal account users who grant Gemini access to these apps should double-check both settings.
We have a full guide on how to turn off Gemini in Gmail and Google Docs if you want to go further with that.
Quick Comparison: ChatGPT vs. Claude vs. Gemini Privacy Settings
| ChatGPT | Claude | Gemini | |
|---|---|---|---|
| Training on by default? | Yes, Free/Plus/Pro | Yes, since August 2025 | Yes, personal accounts |
| Safe by default? | No | No | No |
| Where to opt out | Settings, Data Controls | Settings, Privacy | myaccount.google.com |
| Toggle name | “Improve the model for everyone” | “Help improve Claude” | “Keep Activity” |
| Chat history stays after opt-out? | Yes | Yes | No, both are linked |
| Data kept after opt-out | Up to 30 days (safety) | 30 days | 72 hours |
| Enterprise protected by default? | Yes | Yes | Yes, Workspace managed |
What You Should Never Type Into Any AI Chatbot
Even with every privacy setting off, some chats can still be kept for safety monitoring and legal reasons. So I never put these into any AI tool:
- Full names paired with home addresses or phone numbers
- Passwords or API keys (if you’ve pasted one, change it now)
- Patient or client data, which is a professional and legal issue, not a privacy-toggle one
- Confidential business plans or unreleased product details
- Banking details or full account numbers
Think of a chat window as a whiteboard in a shared office: if the information would hurt someone if seen, it doesn’t belong there. Read the full guide: What You Should Never Type Into Any AI Chatbot
When These Settings Will Not Fully Protect You
There are real limits here, and being honest about them matters.
The first limit is that past training cannot be undone. Data already used in a completed training run is embedded in a model’s weights. Removing it is technically described as “machine unlearning,” and it remains an unsolved problem at this scale. Opting out stops future use. It does not reach backward.
The second limit is that safety monitoring continues regardless of your settings. All three platforms may retain and review conversations that are flagged for harmful content, safety violations, or legal reasons. This is confirmed in their public documentation and is separate from model training opt-outs.
The third limit is that opt-outs are scoped to your account. If someone else pastes your document into their own ChatGPT with training enabled, your data is in their session. Your opt-out does not follow your information. It only governs what happens in your own signed-in session.
The fourth limit is that policy defaults can change. Anthropic’s shift in August 2025 is the clearest proof of this. A company that built its reputation on privacy-first data handling made a significant policy change, and users who had never revisited their settings were affected. Set a reminder to check these settings every six months. Defaults change, and not always with a loud announcement.
The Next Question You Are Probably Asking: Can AI Collect My Data Without Consent?
Technically, no. All three platforms publish privacy policies and provide opt-out controls. Consent is disclosed in their terms.
In practice, though, consent buried in a lengthy terms document that most people never read is not the same as informed consent. The settings are real, but they are not easy to find, and the defaults favor the companies, not the users.
If you are in the European Union, the General Data Protection Regulation gives you stronger rights, including the right to request access to your data, to object to processing, and to request deletion. In the United States, protections vary by state. If you have specific concerns about how your data has already been used, checking your country’s data protection authority is a good next step. In the UK that is the ICO. In the EU it is your national DPA. In the US, the FTC handles consumer data complaints at a federal level.
Your 5-Minute Privacy Checklist
Open each of these in a separate browser tab right now and work through the list before you close this page.
ChatGPT
- Settings, then Data Controls, then turn off “Improve the model for everyone”
- Also turn off voice and video training toggles while you are there
- Visit the OpenAI Privacy Portal and select “Do not train on my content”
- Use Temporary Chat for anything sensitive going forward
Claude
- Settings, then Privacy, then turn off “Help improve Claude”
- Or go directly to claude.ai/settings/data-privacy-controls
- Use Incognito Chat for your most sensitive conversations
Gemini
- Go to myaccount.google.com, then Data and Privacy, then Gemini Apps Activity
- Or go directly to myactivity.google.com/product/gemini
- Turn off Keep Activity
- Delete past activity while you are on the page
- Check Gmail and Docs Smart Features settings separately if you use Gemini there
Frequently Asked Questions
Does ChatGPT Plus use your data for training?
Yes, unless you turn the setting off. The Plus subscription buys speed, feature access, and higher message limits. It does not automatically opt you out of training. Go to Settings, then Data Controls, and turn off “Improve the model for everyone” manually.
How long does Claude keep my conversations if I opt out?
Approximately 30 days, according to Anthropic’s retention documentation. If you opt in to training, that extends to up to five years in de-identified form. Deleted conversations are not used for training under any circumstance, but you need to delete them separately from opting out of training. The two actions are independent.
What happens when I turn off Gemini’s Keep Activity setting?
Your new conversations stop being stored in your account history and stop being used for model training or human review. Google still keeps those conversations for up to 72 hours for service and safety processing before clearing them. Unlike ChatGPT, Gemini links history and training in one setting, so turning it off also removes your saved conversation list.
Can I opt out and still use all the features normally?
Yes, across all three tools. The opt-out does not restrict functionality. ChatGPT answers the same way, Claude works the same way, and Gemini performs identically. Memory and personalisation settings are separate and unaffected by the training toggle.
Is it possible for AI to collect data without my consent?
Not technically, since all major platforms disclose data use in their terms. But consent buried in long policy documents is not always meaningful in practice. If you are in the EU, GDPR gives you stronger rights to object, request access, or demand deletion. In other regions, protections vary. If you believe your data has been misused, contact your national data protection authority.
One Last Thing Before You Go
The settings you just changed are a good start. But they are not permanent. Privacy policies evolve, company strategies shift, and what is opt-out today could change tomorrow. Anthropic’s August 2025 shift from opt-in to opt-out is the clearest example of that in recent years.
Bookmark this guide. Set a calendar reminder for six months from now to revisit your settings on all three platforms. It takes five minutes each time, and it is the single most practical privacy habit you can build around AI tools right now.
If you want to get more out of these tools while keeping your data private, here are a few places to start. Our ChatGPT for teachers guide covers responsible use in educational settings. If you are using Claude for document work, our Claude AI resume prompts guide includes examples built around privacy-conscious workflows. For a full breakdown of what ChatGPT Plus actually gets you, see our ChatGPT Plus subscription guide. And if you are actively trying to remove AI features from Google’s other tools, we have a step-by-step guide on how to turn off Gemini in Gmail and Google Docs.
All steps and policies in this article are based on verified, publicly available information as of October 2026. AI privacy policies change. Revisit each platform’s official privacy documentation after major product updates to confirm nothing has shifted.
