Not to mention potential misbehavior by rogue elements, even when official policy is innocent. LOVEINT being a recent example in precisely this domain, but instances are manifold.
Another important point about LOVEINT in particular is that the known cases were all self-reported. Which speaks to the integrity of those involved compared to hypothetical officers who did the same thing but did not report themselves... but does not speak well of the internal checks in the system, and I have a hard time believing those other officers are entirely hypothetical.
