HTTPBuilder will use the platform default charset for parsing the response. I suppose this could be problematic if a Windows platform's default encoding is iso-8859-1 and you happen to be parsing a non-latin response. Same might apply for OSX platforms (I think the default charset there is MacRoman or something?)
I suppose my default response would be "if you don't see any problems, don't worry about it" If everything looks OK then you can just disable warning logging for ParserRegistry.
I know that's not terribly helpful so I'll explain a bit more. Servers should send a charset as part of the Content-Type header in the response. The warning you see indicates that the server didn't provide a charset, so the parser is using the JRE platform default charset. I'm actually not sure if the HTTP spec prescribes how a response should be interpreted if a charset isn't specified.
The safest bet would be to use UTF-8 if a charset isn't given in the response. Maybe I should change ParserRegistry so that you can set a default charset for such cases. Then you can specify the behavior. I'll leave this issue open for such a feature post-0.5.0.