Nevermind, it has nothing to do with encoding.
Even if I force it bye stating:
string.encode().decode('utf-8')
It throws the same error.
It looks like the problem were the too many \n\t inside the JSON.
A simple “cleaning” strategy like:
import json
new_json = json.dumps(json.load(old_json))
solved the problem.
So the problem was the remote party’s ability to manage carriage returns and/or line feeds in the received JSON string.