mekat.
Try a lesson

How to stop AI training on your data

Most big AI apps let you switch off training on your chats: in ChatGPT it’s “Improve the model for everyone”, in Claude it’s “Help improve our AI models” under Settings, then Privacy, and in Gemini it’s “Keep Activity”. The switches only work going forward, they don’t delete anything, and a few things, like thumbs-up feedback, can still be used, so it’s worth knowing what each one really does.Updated October 25, 2026

Why your data is used for training at all

AI models learn from huge amounts of text, images and conversations. Many companies also use what people type into their chatbots, and what people post on their platforms, to make the next version better. That isn’t secret, but the defaults aren’t always what you’d pick, and the settings live in a different place in every app. The good news: each one takes about a minute to change.

Two things are true almost everywhere:

  • Turning training off applies to new chats and content. A model that has already been trained can’t “unlearn” what it saw.
  • Turning training off is not the same as deleting. Your chat history usually stays until you delete it yourself.

ChatGPT

OpenAI calls the setting “Improve the model for everyone”. It says data sharing is on by default for personal ChatGPT Free, Plus and Pro accounts. When it’s off, OpenAI says your new conversations won’t be used to train its models, and they can still appear in your chat history.

  1. On the web, open your account menu and select Settings.
  2. Select Data controls.
  3. Select Improve the model for everyone, turn it off, and select Done.

In the iPhone or Android app, open the sidebar, tap your profile icon to open Settings, then Data controls. When you’re signed in, the choice follows your account to every device.

Good to know:

  • Voice mode has its own, separate control for sharing data for training, so check that one too if you talk to ChatGPT out loud.
  • Temporary Chat is a one-off private mode. OpenAI says those chats don’t appear in your history, aren’t used to improve its models, and may be kept for up to 30 days for safety.
  • Feedback counts. If you tap thumbs up or thumbs down, OpenAI says the whole conversation linked to that feedback may be used for training, even after you opt out.
  • Work accounts are different. OpenAI says that by default it doesn’t train on ChatGPT Business, Enterprise, Edu or ChatGPT for Healthcare workspaces.

Claude

Anthropic lets people on Claude Free, Pro and Max choose whether their chats and coding sessions are used to improve Claude. When it announced the choice on August 28, 2025, Anthropic said it keeps data for five years if you allow training, and 30 days if you don’t.

  1. Select your name in the menu, then Settings.
  2. Select Privacy.
  3. Under “Help improve our AI models”, turn the switch off.

Anthropic says that once it’s off, new chats won’t be used for future training, and it stops using your older stored chats in future training runs, though training that has already started and models already trained keep what they learned. Incognito chats aren’t used to improve Claude, even with the setting on. Two exceptions: conversations flagged for a safety review can still be used for safety work, and thumbs up or down feedback is stored with the whole conversation for up to five years.

Gemini

Google’s setting is called Keep Activity, found under Gemini Apps Activity.

  1. Open Gemini Apps Activity in your Google Account.
  2. Turn off Keep Activity.

Google says that with it off, future chats won’t be used to train its AI models unless you choose to send feedback. Chats are still saved for 72 hours. Read the fine print on one point: Google says chats already reviewed by human reviewers aren’t deleted when you delete your activity, and are kept for up to three years, disconnected from your account. That’s a good reason not to type anything very private into any chatbot in the first place.

LinkedIn, Meta and other social apps

It’s not only chatbots. Social networks can use what you post.

  • LinkedIn. Its setting is called Data for Generative AI Improvement. LinkedIn says that turning it off means LinkedIn and its affiliates, including Microsoft, won’t use your LinkedIn data to improve content-making models going forward, but it doesn’t affect training that has already happened. It says private messages, including InMail, aren’t used. Feedback you send isn’t covered by the switch; LinkedIn points to its Data Processing Objection form for that.
  • Meta. Meta says it trains on information shared on its apps, such as posts, photos and their captions, plus your interactions with its AI features. It says it doesn’t use the content of your private messages with friends and family unless you or someone in the chat chooses to share them with its AI. Its page doesn’t describe a simple on/off switch for public posts, so a practical step is to share posts you’d rather keep out of the mix with friends only, not publicly. Meta also offers a form for requests about your personal information from third parties being used for its AI.

For any other app, search its help center for “model training” or “improve our models”, and check again after big updates. The FTC has warned that it may be unfair or deceptive for a company to quietly change its terms to start using your data in new ways, including for AI training.

Habits that protect you whatever the settings

  • Don’t paste in what you’d hate to see leaked: passwords, ID numbers, bank details, medical records, or other people’s private information.
  • Use the private or temporary mode for one-off sensitive questions.
  • Skip the thumbs up or down on chats that contain personal details.
  • Delete old chats you no longer need. Opting out doesn’t do that for you.
  • At work, follow your employer’s rules and use the work account they provide.

How Mekat helps

Mekat’s Privacy, safety and scams stage has lessons on what never to paste and on chat history, memory and data settings. See the full learning path.

Quick answers

Does turning off AI training delete my chats?

No. In ChatGPT, Claude and Gemini, the training setting and your chat history are separate. To remove chats, delete them yourself.

Can I remove my data from an AI model that’s already trained?

Generally no. Anthropic and LinkedIn both say opting out doesn’t affect training that has already happened, so the switch protects what you share from now on.

Is ChatGPT’s Temporary Chat used for training?

No. OpenAI says temporary chats aren’t used to improve its models and don’t appear in your history, though they may be kept for up to 30 days for safety.

Does thumbs-up feedback count as training data?

It can. OpenAI says the whole conversation linked to your feedback may be used for training, even after you opt out, and Anthropic stores feedback with the whole conversation for up to five years.

Sources

Written by the Mekat team at Martinez Solutions LLC, Glendale, Arizona. Spotted something out of date? support@mekat.ai