The big problem is that to copy the data you have to read it, and to read the data you have to read it! There is no way for the USB stick to tell if the process that is reading the data is showing it to the user or writing it to the hard disk. Ignore the USB stick, it is not the issue, the issue is the data. You encrypt your data and you need to write an application to display the data to the user, anything else that reads the file will just see gibberish. Your application, which knows how to decode the data, will be the only way to view the data.
Of course this can be hacked, what can be made by man can be broken by man, but you have just upped the entrance fee. Then comes the trade off, is your data valuable enough to cover the expense and inconvenience to the user of the custom software and is it too expensive for the thief to hack the application?
Of course the application could be copied so you now need to lock the application to the hardware so that the data becomes useless without it. As a USB stick is just a computer on a chip you could build some custom hardware that is required by the application (such as the decryption built into silicon) so that the application cannot be run without the USB stick.
And that is the best you are going to get.
Did I mention that this could get expensive :)