Anthropic to use Claude chats for training with optout by Sep 28 2025

Anthropic announced Thursday that it will begin using conversations from its consumer chatbot, Claude, to train future AI models, giving users until Sept. 28, 2025 to opt out. The change, rolled out in an update to the companys Consumer Terms and Privacy Policy, affects users of Claude Free, Pro, Max and Code and will be enabled by default for new and resumed chats unless a user explicitly turns it off.

Under the new policy, Anthropic said it will use chat and coding session transcripts to improve model capabilities and safety, including better detection of harmful content and stronger performance on coding, analysis and reasoning tasks. The company said it will apply automated tools and processes to filter sensitive data and will not share information with third parties. Prior to the change, Anthropic said it did not use consumer conversations to train models unless users submitted feedback.

Users signing up for Claude will see a "Help improve Claude" toggle in the signup flow that can be set to off, and existing users will receive a popup labeled "Updates to Consumer Terms and Policies" that allows them to opt out. The company said users can also change the setting later by...