Anthropic has updated its Consumer Terms and Privacy Policy. The company now asks Claude users to decide whether their chats and code sessions may be used to train future models. Existing users must make a choice by September 28, 2025.
What is changing
Until now, Anthropic has not used consumer chat data to train models by default. That policy has changed. New or resumed conversations may be used for training unless the user opts out. Anthropic will also retain data for up to five years for consenting users. The change covers Claude Free Pro and Max and includes Claude Code. Business and government products are not included.
Anthropic says the update will help it improve model safety and reduce false positives in content filtering. The company adds that using real-world chats can make the models better at tasks such as coding analysis and reasoning. Anthropic also says it will apply automated filtering to remove or mask some sensitive data and that it will not sell user data.
Who is affected
This change affects individual users on consumer plans. It applies to free and paid Claude accounts that are used by people and hobbyists. It does not apply to enterprise offerings like Claude for Work, Claude for Education, Claude Gov, or API use via major cloud providers. If a user wants to keep chats out of training, they must opt out before the deadline.
How to opt
New users will be asked to set their preferences during sign-up. Existing users will see a notification and a pop-up that asks them to accept the new terms and set the model training option. Anthropic shows the option as a toggle, and users can change that setting later in privacy controls. If a user does not pick an option by September 28, they will need to make a choice to continue using Claude.
What this means for privacy
If a user accepts training, then new or resumed chats can be kept and used for model development for up to five years. If a user opts out, the company says those chats will not be used for training. Users should note that data already used for training cannot be removed retroactively. Consumers who worry about sensitive information should avoid sharing it in any chat they do not want stored.
Why Anthropic may have changed course
AI companies need large amounts of real user data to refine models. That data is particularly valuable for safety systems and for improving coding and reasoning skills. Anthropic describes the change as a user choice that will help the product improve. According to industry observers, the move is also an indication of competition and a legal reality of the AI market. Other companies have recorded similar changes in recent months.
