Summary: Anthropic updated its Consumer Terms and Privacy Policy on August 28, 2025 so Free, Pro, and Max users must choose whether chats may train models and support safety work. Opt-in sessions can be retained up to five years, while opt-out users keep the existing 30-day retention. The article says builders must account for this split because consumer-account data assumptions now differ from commercial and API terms.
Anthropic is reworking the way consumer data can feed model improvement. On Aug 28, 2025, the company introduced updated Consumer Terms and a Privacy Policy that hand Claude Free, Pro, and Max users a clear choice: let their conversations be used for training and safety work, or stay on the existing 30-day retention and opt out. If you build on Claude, or ship products that touch consumer accounts, the mechanics of that choice shape your data-handling assumptions.
The opt-in redefines what ‘consumer data’ means for training
Consumer chats previously went unused for training by default. Under the new policy that flips to an opt-in model. Users who agree contribute to model safety and to future Claude capabilities, including coding, analysis, and reasoning. Anthropic says sensitive information is filtered or obfuscated, and that it does not sell user data to third parties.
For builders, scope is the boundary that matters. These updates cover consumer plans only — Free, Pro, and Max — along with Claude Code sessions tied to those accounts. They leave Commercial Terms services untouched: Claude for Work, Claude for Government, Claude for Education, and API usage, including via Amazon Bedrock or Google Cloud’s Vertex AI. Integrate Claude through an API or an enterprise agreement and your data handling stays as it was.
The Oct 8 deadline creates a decision point you cannot skip
Existing users get until October 8, 2025 to accept the updated Consumer Terms and record their training-data choice. Accepting now brings the new policies into effect immediately for new or resumed chats and coding sessions. Once the deadline passes, users must pick a training preference before they can keep using Claude. New users choose during signup.
That deadline lands on product teams that onboard users into Claude-powered features. If your app leans on consumer Claude accounts, your users may meet a pop-up or notification prompting them to decide. The choice changes what data you can expect to persist: opt-in users get a five-year retention window on new or resumed sessions, while opt-out users remain on the existing 30-day retention. Deleted conversations are never used for future training, whichever setting is in force.
What it means for your data pipeline assumptions
If you build tools that process or store Claude conversation logs from consumer accounts, retention is now a concrete variable. A user who opts in has data retained for up to five years — which can touch compliance reviews, data deletion requests, or downstream analytics you run over those logs. For opt-out users the 30-day default is a far shorter window.
This is a different problem from the one Cloudflare addressed with its Disallow AI Training setting, where control over crawler use of content sits with website owners. Here the control sits with the end user, not the site or app owner. Builders who proxy consumer Claude accounts need to surface that control plainly rather than bury it.
The safety angle is part of the tradeoff
Anthropic presents the opt-in as a route to better safeguards against scams and abuse, not merely better model quality. Participants help harmful-content detection become more accurate and less prone to flagging harmless conversations. For consumer-facing products that is a genuine benefit, since false positives from safety filters erode user trust.
The tradeoff is longevity. Five years of retained conversations is a long time for personal data, filtering and obfuscation notwithstanding. Builders who advise users on privacy settings should say it outright: opting in means chats may train future models and be kept for five years unless deleted; opting out preserves the 30-day window but forfeits any contribution to model improvement.
What to watch next
The supplied source does not say how the opt-in rate will affect model quality or safety improvements, and it does not detail the exact filtering and obfuscation methods. Those details matter to builders who want to quantify privacy risk. For now, the practical step is to check whether your product touches consumer Claude accounts and, if it does, prepare your users for the Oct 8 decision point. The setting lives in Privacy Settings and can be changed at any time — but the deadline forces an initial choice.