Program languages can be Turing equivalent however an alternative view of language is its like any other data encoding and via some kinds of Shannon's Law for human speech your language has to be able to achieve a certain SNR equivalent minimum to get an idea across in a small slice of a lifetime devoted to a book at a reasonably low enough error rate to claim a meeting of the minds on a topic. So you can write Kant in German and maybe its possible to understand an English translation of Kant with a bit more effort but there exists a hard lower SNR limit below which it doesnt work such that a Lojban translation of Kant is simply a fliba. Note that its difficult to find a Lojban-English dictionary that isn't also an English thesaurus in practice. There are indications Kant could be translated into Esperanto but nobodies tried in public, or maybe because its also quite impossible.

Now as a counter example (or is it?) see the whole concept of "a quran isn't a quran unless its written in arabic at which point its a real quran" vs the similar fun the pre-reformation christians had with the same belief WRT the new testament.