I've been racking my brains over this one -- it's a tricksy little question! -- and I'm still not sure what the answer should be. Of course Nicholas Smith would be correct, if the question was about the proposition that I am being systematically deceived. But it isn't. I take it that the question is how we know that it is possible that I am being systematically deceived.
Admittedly, Descartes himself does ultimately conclude that this isn't even so much as possible: but he reaches this conclusion via a rather idiosyncratic and unconvincing argument, resting on the nature of God; and, in any case, even he acknowledges that it certainly does seem to be possible. He sets up his methodological scepticism in the First Meditation (as I'm sure you know), pointing to things like optical illusions, dreams, and the possibility of an evil demon. Many of the same points could be made about each of these arguments: but, for simplicity's sake, I shall just take the one about illusions.
So, for instance, he says that, when we look at a tower in the distance, we might take it to be round. But we then get closer, and we now find it to be square. So it turns out that we've been in error in at least one of our judgments, presumably the earlier one. But then how can we be so sure that we are not similarly in error in our current judgments too, and not only here but right across the board? The thing that is driving the argument is the fact that a certain object is appearing to be round, but the very same object is also appearing to be not-round. And it would be a contradiction for one and the same thing actually to be both round and not-round. (We certainly know that a priori). So we can conclude that some of our perceptions or judgments must be false; and that's what opens the door to global scepticism. If some of them are false, then perhaps all of them are false.
But is it really that simple? The tower looked round at time t1, and it looks square at time t2. But maybe it was round at t1, and is square at t2. How can we be so sure that it hasn't simply changed with the passage of time? If we can't be sure of that, then we can't be sure that either of these judgments was erroneous after all. And, if we can't be sure that any of our judgments are erroneous, then it's not clear how we could know that any of them even could be erroneous. And, if we can't know that, then it seems that perhaps we can't know, after all, that it's possible that all of them could be erroneous together. If we can't establish the reality and (a fortiori) the possibility of error, then we'll have nothing to extrapolate globally to establish the possibility of systematic deception. Now, for my part, I do feel entirely confident that we can indeed know that error is possible. But what I'm much less confident about is precisely how we can know this. Simple sense-perception alone certainly won't be enough to establish it.
Well, that's about as far as I've got. We can only know that systematic deception is possible, if we can know that error is possible. That is to say, we can only know that all of our beliefs might be false, if we can know that some of them might be false. And we could know that some of them might be false, if we knew that some of them actually were false. But how do we know that? If I'm forced to give an answer, I guess I'd plump for 'empirical' over 'a priori'. But, ultimately, I just don't know what to think about this. Nice question!