This is all still playing out in the U.S. courts, ultimately, but to date the courts (and thus the Copyright Office) seem to have been holding that claim to copyright originates with human creativity, and that works generated by software have no standing for copyright claim without substantial human creative choice being a part of the process. A machine translation from one text to another wouldn't seem to qualify. (But, IANAL either.)
IANLAL, but as I mentioned earlier, U.S. law has an almost inescapable
bias towards applying copyright to everything, despite the wishes of its
creator. I believe that if one were to use a tool such as AI to perform
a translation, the copyright holder would be the one using the tool, not
the one owning the tool.
I believe that if you were to use Google Translate to create a
translation, you would have a new copyrighted work, and you would be the
copyright holder.