ECMA-262 does not define how to parse a file into a module, when it's imported into JS. For browsers, this is defined by the HTML standard, which includes this:

Let sourceText be the result of UTF-8 decoding bodyBytes.

That applies for all imports, be it for JS modules, JSON modules, or the upcoming text modules.

For text modules, the intent in its HTML standard changes is to match the behaviour of Fetch's .text(), which already now mandates using the UTF-8 decode algorithm for all files.

I'd be surprised if some runtime wasn't currently at least detecting a BOM and potentially parsing UTF-16 content as UTF-16 rather than UTF-8.