Anthropic on Thursday announced plans to collect Claude chat data to train future versions of the AI chatbot, giving users the ability to choose whether to have their chats included in that training data. That’s the usual behavior for generative AI products like ChatGPT and Gemini. The chatbots are trained on lots of data sources during development, including information from conversations users might have with the AI. It’s actually surprising that Anthropic did not collect chat data in previous years, considering that OpenAI and Google have the setting turned on by default. You have to opt out to avoid your data being used to train future versions of ChatGPT and Gemini.
The same privacy setting is now available to Claude users, including current and new customers. You’ll be able to opt out of Anthropic’s data collection if you choose to, and you can change your mind at any time. You might want to help Anthropic develop better Claude versions; there’s nothing wrong with that. If you want your data to be part of future Claude training runs, you should still be mindful of what information you give the chatbot during your conversations and avoid sharing data that’s too personal.
Anthropic’s privacy change applies to Claude Free, Pro, and Max plans. These are the only Claude accounts that will be impacted. Claude for Work, Claude Gov, Claude for Education, and API use will not be impacted, meaning these Claude accounts will not contribute data for model training in the future.
How to stop Claude from training on your chats
Anthropic won’t turn on data collection automatically for new and existing users. The company explained in a blog post that starting Thursday it will send out notifications to existing Claude users to review the new privacy settings. They will have until September 28 to accept the new terms of service and choose whether to have their data train future versions of Claude. Anthropic says that accepting the changes when you receive the notification means the privacy update will apply immediately to new and resumed chats or coding sessions. Once the deadline passes, you’ll need to make a choice to keep using Claude.
Stopping Claude from training on your chats is very easy. Once you see the notification above, disable the “You can help improve Claude” option. That means toggling off the setting. Then tap “Accept” and proceed with your Claude chats. The toggle will be available in Claude’s Privacy Settings if you change your mind in the future, or if you pressed that Accept button too fast.
If you’re a new Claude user, you’ll be shown the setting for model training during the sign-up process. Like existing users, you’ll be able to change your mind later.
Since Claude users might also use rival products like ChatGPT and Gemini, you should know you can stop ChatGPT from training on your chats. Similarly, you can prevent Gemini from training on your personal data. You should check the privacy settings of any other chatbot you might use for similar options.
How Anthropic stores and uses your data for Claude training
The company also announced that it’s going to retain your data for five years if you choose to help Anthropic train Claude. The new data retention policy applies to new and resumed chat sessions. If you choose to stop Claude from training on your personal chats, Anthropic will retain your data for a 30-day period. Also, Anthropic will retain chats for five years if you submit feedback on interactions with Claude.
If you delete a Claude chat, that conversation won’t be used for model training in case you let Anthropic collect your chats for improving the model. If you turn off model training in the future, Anthropic will stop collecting data from Claude interactions that occur after you’ve disabled the feature. Any chats you allowed Anthropic to use for Claude training will continue to be used for model training runs that have already started. But Anthropic will not use that data for future training runs.
Importantly, Anthropic also explained that it uses “a combination of tools and automated processes to filter or obfuscate sensitive data” if you allow your data to be used for Claude training. Still, you should avoid giving sensitive information to any chatbot. Finally, Anthropic says that it does not sell users’ data to third parties.
