@OliviaVespera I can't tell where they source their training data. So, I'm guessing not.There's a general reference to "Opus", but there's a lot of sets of training data on Opus likely with various copyright holders. Opus site uses the word "open" a lot, but doesn't say how they verify the data is freely redistributable, what terms, or how they communicate with the authors / copyright holders.That would be my main _ethical_ concern.I also can't vouch for the accuracy, and I would be concerned that they don't provide ANY source of metrics for how accurate the translations are. (I don't how enough about machine translation to how if there's a standard metric for accuracy or not.)
