Anthropic's commercial products — including Claude for Work, Anthropic API, and Claude Gov — do not use customer inputs or outputs for model training by default. However, when users explicitly provide feedback via the thumbs up/down button, the entire related conversation is stored in secured back-end systems for up to 5 years. This feedback data includes conversation content, custom styles, preferences, and model settings, but excludes raw content from connectors like Google Drive or MCP servers unless directly copied into the chat.
Anthropic de-links feedback from user and customer IDs before internal use. The data may be analyzed for service effectiveness, research, user behavior studies, and AI model training as permitted by law. Feedback is not combined with other user conversations.
Organization Primary Owners or Owners on Team or Enterprise plans can disable the feedback submission feature entirely through the Rate chats setting under Organization settings > Data and Privacy. For full details, users should consult Anthropic's Trust Center and Commercial Terms.
Source: Hacker News · Summarized by HeadlinesBriefing