Anthropic May Train on Rated Chats and Keep Them Up to Five Years
Anthropic's policy says thumbs-up/down feedback may be used to train models and is stored with the full conversation for up to 5 years, though IDs are de-linked and raw connector data is excluded unless copied in.
Anthropic says it will not train on inputs or outputs from commercial products such as Claude for Work and the Anthropic API by default. If a user explicitly reports feedback or bugs through the thumbs-up/down button, the company may use the related chats and coding sessions to train its models.
The policy says feedback is stored in a secured back end for up to 5 years, de-linked from user and customer IDs before use, and not combined with other Claude conversations. Raw content from connectors such as Google Drive and remote or local MCP servers is excluded unless copied directly into the conversation.
Team and Enterprise owners can disable member feedback through the Rate chats setting.
Key takeaways
- Feedback data includes the entire related conversation, along with custom styles, conversation preferences and model settings, not only the rated message.
- Anthropic says it may use feedback to analyze service effectiveness, conduct research, study user behavior and train AI models as permitted under applicable laws.
- The policy distinguishes commercial offerings from consumer plans such as Claude Free, Pro and Max, which have separate data terms.
What people think
Analysis details
5 comments analyzed. Analyzed .
The sample selects active discussion branches and includes available parent comments. It may omit parts of the full thread. The analysis included 5 of 7 usable stored comments because of the input limit. Selected themes do not measure community opinion or how common a view is.
Feedback retention and training scope
Anthropic's policy says thumbs-up/down feedback stores the entire related conversation for up to 5 years and may be used to train models. The de-linking from user and customer IDs and the promise not to combine feedback with other conversations are cited as mitigations.
Claude Code feedback boundaries
Whether dismissing feedback or answering the 'how is Claude doing' prompt counts as feedback under the policy remains unclear. The policy's examples focus on thumbs up/down, leaving other Claude Code interactions open to interpretation.
Dismissal claim rests on speculation
The assertion that dismissing feedback in Claude Code is treated as feedback rests on a linked write-up and is disputed as speculation; the behavior should not be treated as established without better evidence.
Promises and self-censorship
The gap between explicit sign-up promises and feedback-based training can feel like a breach, and perceived surveillance can lead to self-censorship and lower productivity.
Rating as explicit signal
Rating a conversation can be read as an explicit signal that the chat may be reviewed or used for training, which makes the policy unsurprising on that reading; that reading leaves open whether dismissal or other UI interactions count.