Anthropic’s commercial-product guidance says Claude for Work and API inputs and outputs are not used for training by default, but explicitly submitted feedback, bug reports or separate permission are exceptions. Thumbs-up or thumbs-down feedback can include the entire related conversation and settings, retained for up to five years. Raw connector and MCP server content is excluded, but material copied directly into the conversation may be included. Team and Enterprise owners can disable feedback submissions through Rate chats under Organization settings, Data and Privacy. Free, Pro and Max accounts follow a separate consumer policy.