User blog comment:IamThorn/Can Penny be hacked?/@comment-24891101-20141123054155

People have already dealt with the issue of requiring an external connection for remote hacking to occur, so I shall dispense with that line of inquiry for the most part, save to note that her stated role is intended to be that of human oversight for the robotic forces of Atlas. This position requires her to be tied into the C2 system, and why not have it onboard, since the system is in use already and therefore presumably secure? However, all that would do is allow the hacker to spoof her knowledge of where assets are, not take control of her, since that's beyond the ability of the system.

On a different note, however, Penny's neural net is likely not the be-all and end-all of her own systems, as you alluded to. She's got all sorts of embedded systems which control her propioception and actualize her movements, borrowed from the AK series; so too for sensory stuff. Those are probably also isolated.

Far more interesting is the possibility of a failsafe AI. Since Penny is a testbed, the behavior of which her creators would not know, it's quite possible that she's outfitted with an onboard AI she doesn't have access to, that can be remotely activated in the case that she runs amok, which would interdict her control over her body, and possibly be able to puppeteer it. That could be compromised. This AI would be based on existing ones, and have loyalty programming.

The last, most prosaic, and most worrying possibility is that she could simply be ordered, via her C2 system, to kill people. Since Black Queen, if it were intended to compromise Atlas' command network, would simply involve giving someone credentials they shouldn't have, Penny would see the command authorization, and might be naive enough to follow through, not realizing the order is probably falsified.