Hi Vasco,
indeed the CSV files do not have magic numbers, and the possible outputs are always 'Text/Plain' or 'Application/octet-stream', depending on the file size.
These two outputs reveal that the file content is ambiguous. They mean exactly the same. More info can be found here.
Your best shot would be to perform the following validation:
(Extension = '.csv' and (mimetype = 'text/plain' or mimetype = 'application/octet-stream'))
If this condition is true, then most likely the file is a csv file. This is not bullet proof but since you have no magic numbers available, it's your best shot.