Wispr Flow
Wispr
Careful
Last checked 10 Nov 2026 · Reviewed by Aaron Simmons, security practitioner · How we decide
Wispr says it uses your audio, transcripts, and edits to train models only if you allow "Improve the model for everyone," and that any user can turn that off. On Growth and Enterprise the control is off by default and locked. If dictation cloud storage is off, audio is discarded once it is transcribed. Notetaker audio is kept encrypted for a short period that is not given as a number of days. The pages we read do not describe a voice-cloning product. A personal account is Careful until you confirm that training control.