TLDR: Anthropic, the developer of the Claude AI chatbot, is updating its privacy policy to use user chat transcripts for training its AI models. This change, effective September 28, 2025, will affect users of Claude’s free, Pro, Max, and Code tiers. Users must actively opt out through their settings or during sign-up, otherwise their data will be used for training and retained for up to five years. Enterprise and API users are exempt from this policy shift.
Anthropic, a prominent AI startup backed by Amazon, has announced a significant revision to its Consumer Terms and Privacy Policy, set to take effect on September 28, 2025. The updated policy will allow the company to utilize chat and coding session transcripts from its Claude AI chatbot to train and enhance its artificial intelligence models. This move marks a notable shift from Anthropic’s previous stance, where user conversations were not used for training unless explicitly submitted as feedback or opted into.
The new policy will impact a broad spectrum of Claude users, specifically those on the Claude Free, Pro, Max, and Code subscription tiers. For these users, inaction will be interpreted as consent; if no explicit opt-out choice is made by the September 28 deadline, their chat data will be automatically included in the training datasets.
However, certain user categories are exempt from this policy change. These include users accessing Claude under specific commercial licenses such as Claude for Work (Team and Enterprise plans), Claude Gov, and Claude Education. Additionally, third-party users who integrate with the Claude API via platforms like Amazon Bedrock and Google Cloud’s Vertex AI will also not have their data used for model training under these revised terms.
Anthropic has outlined a clear process for users wishing to opt out. Existing Claude users will encounter a pop-up notification titled ‘Updates to Consumer Terms and Policies’ upon logging in. To prevent their data from being used, users must ensure the ‘You can help improve Claude’ toggle is switched off (typically indicated by being set to the left or an ‘X’ mark) before accepting the updated terms. For new users, this opt-out option will be presented during the initial sign-up process. Users who have already accepted the terms or wish to change their preference later can do so by navigating to Claude’s Settings, selecting the Privacy option, and toggling off ‘Help improve Claude’ in the Privacy Settings.
It is crucial to note that while users can opt out at any time, this decision will only apply to future chats. Any data already collected and used for training prior to opting out cannot be recalled. Furthermore, for users who opt in, Anthropic will extend its data retention period for these chat transcripts to up to five years, a significant increase from the previous 30-day deletion policy for consumer data. For those who opt out, the 30-day retention period will continue to apply, unless data retention is required for policy or legal reasons.
Also Read:
- Anthropic, Backed by Amazon, Unveils AI-Powered Financial Services Platform
- Anthropic Seeks Curriculum Lead to Advance Global AI Education Initiatives
Anthropic states that the rationale behind this policy update is to foster the development of ‘even more capable, useful AI models’ and to ‘strengthen safeguards against harmful usage like scams and abuse.’ This change comes amidst a broader industry trend where AI companies are increasingly updating their privacy policies to leverage vast amounts of user data, which is considered vital for the rapid advancement of generative AI technologies. The move has already sparked discussions regarding user privacy, transparency, and the nature of informed consent in the evolving landscape of artificial intelligence services.
