I suppose that the likelihood estimate would need to take into account the number of people who have (or had) access to the sources. Obviously the alternatives are not mutually exclusive.