A list of credit card numbers could be modified so that, for example, real people are not affected, and/or to add "watermark" card numbers that identify the precise origin of the dataset if it is ever used.
Hopefully there are strong guards against modified data being used to entrap, and also against modified data being used as evidence.
(I'm not yet sure whether these policies are sensible; just trying to figure out the legitimate uses for these powers rather than assuming the worst, although the latter can also be useful to prevent bad outcomes)