Does ChatGPT Train on Your Chats?

Last checked: Nov 8, 2026. Based on OpenAI Help Center articles and Privacy Policy dated Feb 6, 2026.

If you use ChatGPT on a personal Free, Plus, or Pro account, what you type may be used to train OpenAI's models unless you change a setting that's on by default.

That includes work notes, medical questions, and source code people paste without thinking twice.

Here's what data is used for training, what's protected, and how to turn it off.

#What ChatGPT uses for training

On personal ChatGPT Free, Plus, and Pro workspaces, data sharing is enabled by default (OpenAI Help Center, 2026).

That means your conversations may be reviewed by OpenAI's AI trainers and used to improve the models, according to OpenAI's Help Center article "Data controls in ChatGPT" (2026).

#What may be used

When data sharing is on, OpenAI may use:

• Your prompts (what you type into ChatGPT)
• ChatGPT's responses
• Files you upload or create in conversations
• Feedback you provide (thumbs up or down)

The data may be reviewed by human AI trainers or used in automated training processes.

#What's not used for training by default

Business, Enterprise, Edu, Healthcare and API workspaces are not used for training by default (OpenAI Help Center, 2026).

Temporary Chat conversations are not used to improve models while they remain temporary. If you save a Temporary Chat, it becomes a regular chat subject to your account settings. Temporary chats are normally deleted within 30 days, subject to safety and legal retention exceptions (OpenAI Help Center, 2026).

#The Samsung incident: what happened in 2023

In May 2023, Samsung banned ChatGPT and other generative AI tools on company devices after employees uploaded sensitive internal source code to ChatGPT (The Verge, May 2, 2023, citing Bloomberg).

According to Fortune (May 2, 2023), a Samsung engineer uploaded confidential source code while using ChatGPT to help check the code.

Samsung's concern was that the material was now on outside servers and would be difficult to delete (The Verge, May 2, 2023).

The incident does not mean Samsung's code later appeared in someone else's ChatGPT answers. The risk was that the data had left Samsung's control and was stored on OpenAI's servers.

#What "may be used for training" means

When OpenAI says conversations may be used to train models, that does not guarantee every chat is reviewed or trained on.

It means OpenAI reserves the right to use the data. Some conversations may be sampled for training, while others may not be used at all.

According to OpenAI's Privacy Policy (Feb 6, 2026), personal information collected may be used to develop and improve services, which can include training AI models.

#How to turn off data sharing in ChatGPT

To prevent your new conversations from being used to train OpenAI's models:

1. Open ChatGPT (web, desktop, or mobile).
2. Click or tap your account menu (profile icon).
3. Go to Settings.
4. Open Data controls.
5. Find "Improve the model for everyone" and turn it off.
6. Click or tap Done.

#What happens to old conversations

Turning off data sharing does not delete saved chats (OpenAI Help Center, 2026).

The opt-out applies to new conversations. Content already used to improve models may have been de-identified and disassociated from your account.

#Thumbs up and down still send data on personal plans

On personal Free, Plus, and Pro accounts, giving a thumbs-up or thumbs-down on a ChatGPT response can still send that whole conversation into the training pipeline, even with data sharing turned off (OpenAI Help Center, 2026).

Feedback handling may differ on managed Business, Enterprise, and Edu workspaces.

If you don't want a conversation reviewed, avoid using the thumbs-up/down feedback buttons on personal accounts.

#What to do if you've already shared sensitive information

If you've pasted work documents, source code, or personal details into ChatGPT while data sharing was on, here's what you can do:

#1. Turn off data sharing now

Follow the steps above to prevent new conversations from being used for training.

#2. Delete old conversations

Go to your ChatGPT history and delete conversations that contain sensitive information. Deleting a chat removes it from your account, but OpenAI states that content already used to train models may have been de-identified.

#3. Use Temporary Chat for sensitive topics going forward

Temporary Chat conversations are not used to improve models while they remain temporary. If you save a Temporary Chat, it becomes a regular chat subject to your account settings.

Temporary chats are normally deleted within 30 days, subject to safety and legal retention exceptions.

Look for a Temporary Chat option in your ChatGPT interface to start a chat that won't be used for training.

#4. Consider a Business or Enterprise workspace for work projects

If you use ChatGPT for work, ask your employer about a Business or Enterprise plan. Those workspaces are not used for training by default.

#Why this matters for privacy

When you paste text into ChatGPT, you may include details that reveal more than you realize:

• File paths and folder names that show your company or project structure
• Variable names that reference internal systems
• Comments in code that mention coworkers, clients, or deadlines
• Error logs that include server names or IP addresses
• Medical or legal questions that describe a specific situation

#Understanding the scope of data controls

The "Improve the model for everyone" setting covers ordinary personal text chats in ChatGPT on personal Free, Plus, and Pro accounts.

However, OpenAI's data controls may not cover every training channel. According to OpenAI's Help Center, there may be separate controls for voice conversations and other specialized environments. Some session types, such as Trusted Access or Daybreak sessions, may have different policies.

Turning off data sharing does not guarantee that no aspect of your ChatGPT use will ever contribute to model improvement. It controls the primary training channel for standard text conversations.

#Protecting your data beyond ChatGPT

AI training policies are just one part of online privacy. Your personal information is already listed on data brokers and people-search sites that anyone can access.

Those sites collect your name, address, phone number, and relatives from public records, marketing lists, and old breaches, then sell access to that data.

Even if you opt out, many data brokers re-list you after a few months, so you have to check and remove your listings over and over.

UndoLeak erases your data from data brokers and keeps it off.

If you want to see what personal information about you is already listed online, run a free scan.

#Summary

On personal ChatGPT Free, Plus, and Pro accounts, data sharing is enabled by default. Your conversations may be used to train OpenAI's models unless you turn that setting off.

Business, Enterprise, Edu, Healthcare and API workspaces are not used for training by default. Temporary Chat is not used for training while it remains temporary.

To opt out:

1. Go to ChatGPT → account menu → Settings → Data controls.
2. Turn off "Improve the model for everyone".
3. Click Done.

Turning off data sharing does not delete saved chats. Content already used for training may have been de-identified and disassociated from your account.

Avoid using thumbs-up or thumbs-down feedback if you don't want a conversation reviewed. Use Temporary Chat for sensitive topics, or consider a Business or Enterprise workspace for work projects.

#Sources

1. OpenAI Help Center, "Data controls in ChatGPT" (2026)
2. OpenAI Help Center, "How your data is used to improve model performance" (2026)
3. OpenAI, "Privacy Policy" (Feb 6, 2026)
4. Jon Porter, The Verge, "Samsung tells employees not to use AI tools like ChatGPT, citing security concerns" (May 2, 2023)
5. Nicholas Gordon, Fortune, "Samsung threatens to fire employees that leak data to ChatGPT" (May 2, 2023, updated May 4, 2023)