What data does a dictation app send?

Most dictation apps with AI rewriting or cloud transcription send your recorded audio, or at minimum the raw transcript, to a server so it can be processed there. That’s true even for apps that offer a local transcription mode: often the AI formatting or rewrite step still goes through a cloud model.

Dicta’s answer is architectural: speech recognition and the AI rewrite both run on your Mac, using on-device models (Apple’s Foundation Models or a local Qwen3-4B). There is no audio upload path in the app at all. See the Privacy page for the small list of non-content network calls Dicta does make (model download, update check, license refresh, optional anonymous counters).