What Data Does ChatGPT Collect and How to Limit It

ChatGPT has over 300 million weekly users, and most of them write in it like a personal diary. Questions about health, finances, relationships, work… But few know exactly what data ChatGPT collects and what it does with it. In this article I break down the reality behind OpenAI’s data policy and show you how to limit what it shares.

Table of contents

What data does ChatGPT collect exactly

According to OpenAI’s privacy policy updated in 2026, ChatGPT collects the following categories of data:

Data you provide directly:

• Everything you type in conversations (prompts and messages)
• Files you upload: PDFs, images, code, documents
• Account information: name, email, phone number, payment method
• Voice data if you use the voice conversation feature
• Images if you use vision features

Data collected automatically:

• IP address and approximate location
• Device type, browser, and operating system
• Date, time, and duration of each session
• Usage patterns: frequency, types of queries
• Cookie and tracking technology information

Third-party data:

• Social media information if you sign in with Google, Apple, or Microsoft
• Data from business partners and advertisers

What concerns me most personally is the first category: your conversations. OpenAI explicitly admits it uses conversations to train and improve its models, unless you opt out.

Pro-tip: Every time you write something in ChatGPT, ask yourself: “Would I mind if an OpenAI employee read this?” If the answer is yes, don’t write it. Conversations are reviewed by humans to improve the model, and this is in the terms of use.

How OpenAI uses your data

OpenAI uses your data in several ways you should know about:

Model training: This is the most significant one. Your conversations are used to improve GPT-5 and future models. Every prompt you write can become part of training. OpenAI says it applies anonymization techniques, but the real effectiveness of these techniques is debatable.

Service improvement: They analyze usage patterns to understand which features are popular, where the model fails, and how to improve the experience.

Security and moderation: Conversations flagged as potentially dangerous or violating policies are reviewed by human moderation teams.

Legal compliance: OpenAI may share data with authorities if it receives a court order or valid legal request.

Targeted advertising (partially): Although OpenAI claims not to use conversations for advertising directly on its platform, profile and usage data can influence advertising from associated third parties.

How to limit the data ChatGPT collects

Now the important part: what you can do about it. There are several concrete actions you can take:

Disable training with your data

1. Open chat.openai.com and log in.
2. Click on your profile (bottom left corner).
3. Go to Settings > Data Controls.
4. Turn off “Improve the model for everyone.”

With this option disabled, OpenAI won’t use your new conversations to train models. Note: this doesn’t delete old conversations that were already used.

Disable conversation history

1. In Settings > Data Controls.
2. Turn off “Chat History & Training.”

With this, your new conversations won’t appear in the sidebar history and won’t be used to train models. Conversations are automatically deleted after 30 days (according to OpenAI).

Delete existing conversations

1. In the sidebar, click “Clear conversations.”
2. Confirm the deletion.
3. To delete individual conversations, click the three dots on each chat and select “Delete.”

Review and delete memory

ChatGPT has a memory feature that saves data across conversations:

1. Go to Settings > Personalization > Memory.
2. Review what it has memorized.
3. Delete specific entries or all memory.
4. You can disable memory completely.

Export and delete your account

If you want the nuclear option:

1. Go to Settings > Data Controls > Export Data.
2. Request a copy of all your data.
3. Go to Settings > General > Delete Account.
4. OpenAI says it deletes data within 30 days.

Pro-tip: If you use ChatGPT for work, create a business account (ChatGPT Enterprise). Business accounts have data processing agreements (DPAs) that guarantee your conversations aren’t used to train models. It’s the most secure option for sensitive company information.

Differences between free and paid versions in data collection

Many people think paying for ChatGPT Plus or Pro changes data collection. The reality is more nuanced:

Free version:

• Your data is used to train models by default.
• You can manually disable training.
• Access to GPT-4o-mini.
• History saved by default.

ChatGPT Plus ($20/month):

• Same collection options as the free version.
• You can disable training the same way.
• Access to GPT-4o, which can give more accurate responses.
• No difference in privacy policy.

ChatGPT Enterprise:

• Conversations excluded from training automatically.
• Data processing agreement (DPA).
• Centralized administration for businesses.
• Data retention controlled by the company.

ChatGPT Team:

• Similar to Enterprise but for small teams.
• Conversations excluded from training by default.

The takeaway: paying for Plus doesn’t improve your privacy. Only business versions offer real data protection guarantees.

Alternatives if privacy is your priority

If after reading all this you don’t feel comfortable with ChatGPT’s data collection, there are alternatives:

Claude (Anthropic): Similar data policy, but Anthropic puts more emphasis on safety. Claude also has an option to disable training.

On-device AI: Run models locally with tools like Ollama or LM Studio. Your data never leaves your computer. You need powerful hardware.

Gemini with activity disabled: If you already use the Google ecosystem, you can use Gemini with activity disabled. Google still collects some profile data, but not conversations.

Perplexity: Works more like an AI search engine. It collects query data but has reasonable privacy options.

FAQ: Frequently asked questions

Can OpenAI read my ChatGPT conversations?

Yes. OpenAI admits that employees may review conversations to improve the model and for content moderation. This happens even with the paid version. Only business accounts have no-review guarantees.

Is ChatGPT GDPR compliant?

OpenAI says it complies with GDPR and offers rights to access, rectify, and delete data. However, the Italian data protection authority investigated it in 2024 and the legal situation in Europe continues to evolve.

Does deleting conversations completely remove them from servers?

According to OpenAI, deleted conversations are removed from its servers within a reasonable timeframe. But if they were already used to train models, the training can’t be reversed.

Is it safe to write passwords or bank details in ChatGPT?

No, never. Although OpenAI says it applies security measures, your conversations can be accessed by employees and are stored on its servers. Never write sensitive information like passwords, bank keys, or social security numbers.

Conclusion

Knowing what data ChatGPT collects is the first step to protecting yourself. OpenAI collects practically everything: conversations, files, usage data, and profiles. The good news is you have tools to limit it: disable training with your data, disable history, delete memory, and if necessary, delete your entire account. Take 5 minutes today to review your ChatGPT privacy settings. It’s 5 minutes that can protect your personal information.