Hey everyone 
If you use Activepieces in a language other than English, or if you’ve ever contributed translations to a piece, this concerns you.
The automated machine-translation workflow currently shipping with Activepieces is overwriting human-validated translations with low-quality MT output across multiple pieces and languages. The result is action names, trigger names, dropdown options, and field labels that are — in many cases — incomprehensible in the target language, or worse, factually wrong (think statistical operators translated into different mathematical concepts, or HTTP verbs translated as common verbs of the target language).
I’ve personally seen carefully curated translations I contributed get reverted to MT garbage weeks later, with no notification and no review. Talking to other contributors, this is a recurring, structural problem — not an isolated incident.
I’ve opened a detailed issue on the repo with concrete categories of failure and proposed fixes. The core principle being asked: MT is fine when no human has translated a string yet, but a bot must never overwrite human-validated work.
github.com/activepieces/activepieces/issues/13530
If you’ve been affected by this — please add your examples in the issue, or 
it. The more visibility this gets, the harder it is for the team to defer.
Thanks 