Using the ATLanguages plugin seems to be a problem when the text includes some accents. The text is correctly translated ad encoded in base64, but when the text encoded in base64 is transformed to UTF-8, the content shows strange symbols.

This issue is happening because the decoding is shown in ISO-8859-1.

Some findings the customer shared:

These are the class where the base64 change is performed:

info.magnolia.translation.ext.atls.translator.AtlsSynchronousTranslator.
info.magnolia.translation.ext.atls.translator.AtlsAsynchronousTranslator

The exact line:

String decodedBase64 = new String(Base64.getDecoder().decode(base64.getBytes()));

A posible solution:

String decodedBase64 = new String(Base64.getDecoder().decode(text), StandardCharsets.UTF_8);