Mux has published terms that incorporate the AI Services Addendum directly. The public text dated October 12 replaces April’s standalone addendum and applies by default without a separate activation or acceptance step.
Section 3.2 retains the contractual prohibition on third-party AI vendors training their models on customer data or outputs. It distinguishes Mux itself: the company may train, fine-tune or improve its own models using data and outputs that contain no personal data or have been anonymized. The previous text said Mux would not itself train models on data submitted through AI Services, while reserving rights concerning aggregate and anonymous data.
Publishing terms does not establish that every account has already moved to the same contractual regime. Signed agreements may contain exceptions; customers must verify applicable dates and consent obligations. Default applicability of the addendum does not mean an AI video job runs without a request.
“Our vendors do not train their models” does not answer whether the service provider trains its own. The clauses need separate scrutiny.