OpenRouter processes the messages you send to the chat assistant, together with the work data the assistant retrieves to answer them, and work activity and meeting transcripts sent for parts of background processing — grouping related work, extracting decisions and action items from meetings, and interpreting your organization's terminology and goals. Requests are restricted to zero-data-retention endpoints, which do not store prompts or responses after returning an answer, and the data is not used to train models. OpenRouter routes each request to one of a pool of model providers under its own terms; it publishes those providers as a category in its sub-processor list rather than as a fixed set, because routing varies per request.