Just real quick and maybe I'll give you more input later after having thought more about it(this week I have 3 big tests, I'm about to take the 3rd, so I don't have a lot of time right now). After having read The Ender's Game series with Jane having complete control over any computer she wants within her network, I started to think about that. I thought more of an AI, which in my opinion would be infinitely more useful than a human brain tied to a computer, or SYG(tm). For two simple reasons I can think of off hand: Size of Memory and speed of execution.

With the SYG, are you inferring that there would be a processor on both sides? I see that as the only way, and I'm assuming that's how the frankenrobot works, and how some of the new brain sensor periphreal devices work. Where there is a certain signal from the brain, and then the computer processor translates that signal and performs the appropriate action. Of course, I doubt the person would be directly tied to the processor, but rather a software interface between (This is because you want a wide array of processes available to the brain, and most processes performed by processers are basic processes, done a number of times). Say for example you want to move just one pixel across the screen(as as been done with brian sensory equipment), the processor probably takes 100 cycles to do that, it has to add one to the value of the block, however wide the screen is, then has to redraw the screen each time, and store the new value of the block. This would all be done with one thought from the person, but with many 'thoughts' from the processor. Here, you see, it is probably impossible to tie the person directly to the processor. The person would be learning the processor's language, rather than the other way around.

So the software is calibrated to the person, and then he can do a lot with the computer. But how much? He is of course limited by two things. The hardware and software. How complex is the software? If I say in my head that I want to run a brute force hacking system of a network password, how would that be controlled exactly? I am of course limited to the hardware, so if the network cable were say unplugged, I'd be screwed. If the processor blows a fuse and can no longer multiply, I also cannot multiply. Althought such things are unlikely to occur, I'm just trying to let you know how you would be limited by the processor. The software also limits you, though. If you want to run the brute force hack, how would that be handled? Maybe you have complicated software that has that exact command, so when you think it, it happens. Or maybe you have more advanced software, that you can think "I want to throw every possible word in a dictionary at this password, then I want to start trying every combination of letters, number, and symbols" and the software automatically recognizes these thoughts(which of course were calibrated at some point), and does it. But it really isn't that much of an advantage compared to someone typing it, because you are still limited the same as a regular computer in the speed you can execute that hack.

I will say this, one thing the SYG would have over someone else is possibly writing their own software. Since you wouldn't have to repeatedly type and thinking of lines of code would be much faster, then you could essentially program your own software. However, in the end, your software would still would have to adhere to the same network protocols that any normal hacking software would. In short, I don't think SYGs would pose a threat as a "rogue hacker"

Sorry it's so disorganized I just typed as I thought, didn't proofread at all. But those are my initial thoughts on the SYG as a hacking device. I'm not saying it's useless, but that it would prove more useful in other cases... I will post more on it later.

Possibly like you said, learning how the brain works to make a fully function non-biological AI? Now that's a different story altogether.