Anthropic says it will use Claude chats to train AI models: What changes

Claude users face a September 28 deadline to decide if their conversations will train Anthropic's AI, shifting from 30-day deletion to five-year retention

Anthropic has announced changes to how it manages consumer data, giving users of its Claude AI products the option to allow their conversations to be used for training future models. The company said all Claude Free, Pro, and Max users, including those using Claude Code, must make a choice by September 28, 2025, on whether to share data for model development.

 

According to Anthropic, opting in will permit the company to retain conversations and coding sessions for up to five years, with the information used to improve both model safety and performance in areas such as reasoning, coding, and analysis. Users who decline will continue to have their data deleted within thirty days, unless flagged for policy or legal reasons. 

  The update does not affect Anthropic’s enterprise offerings, such as Claude for Work, Claude Gov, Claude for Education, or API access provided through platforms like Amazon Bedrock and Google Cloud Vertex AI. The company said these services follow separate contractual agreements and are excluded from the new policy.

 

For new customers, the choice will be presented during signup. Existing users will receive a notification prompting them to review and update their privacy settings. Anthropic stated that users remain in control, as conversations deleted manually will not be included in training, and clarified that it does not sell user data to third parties.

 

  Anthropic emphasised that the changes are intended to give consumers clearer options while supporting the continued development of its AI systems. The company added that users must make a selection by the September deadline in order to continue accessing Claude.