Suffice it to say that I think that a model that says "collect everything, and let agents analyze it without individualized suspicion, but try to flag questionable searches" is a radical departure from our current approach, which requires justification before collection, and it is also a radical departure from a model that says "collect everything, but require individualized suspicion before analysis."
I don't agree with your assertion that "the risk of rule-breaking is pretty much the same whether the collection comes first or second." If the collection comes first, the data is there to be abused and the risk that it might be abused exists as long as it is there. For example, if the data is stored for 5 years, than at any point in that 5 year period of time an agent decides to make an abusive query, the data will be there waiting for him.
However, if collection comes second, then the data isn't there. It disappears the minute the phone company decides that it is no longer needed for their purposes, and at that point, the risk that it will be abused is over. Any new data created after the point that the agent decides to do something abusive will be available, but the old data will not be. Therefore, the risk to an individual person is significantly lower, and individual people can make real time risk assessments regarding the situation going forward.
For example, lets say you phone Martin Luther King because you support his views on civil rights. Two years later, King makes a public statement opposing a war in Vietnam. At that time, it occurs to you that the military might target King and his associates because King is an influential person who opposes a war that the military wants to get into and they want to shut him up and they want to shut up anyone he might have influenced.
In the collect first scenario, you have to wait for five years after that phone call before you no longer have to worry that the military will find out that you associated with King and target you personally for having done so. In the collect second scenario, you only have to wait six months, or a year at most, until the phone company destroys the data.
The total risk to you in associating with King is significantly reduced the second scenario. Therefore, your likelihood of feeling free to associate with King is significantly greater. Furthermore, if at some point King does something that might make him the target of an illegal government crackdown, you can choose at that point to stop associating with him if thats what you want to do, without having to worry about all of your past associations being exposed.
In other words, the collect first model places a significant chill upon your exercise of your right to freedom of association, because if you ever associate with anyone who later does something that makes him a target of a illegal government crackdown, you could be targeted as well. In the collect first society, you're best off avoiding contact with anyone that might, in the future, be considered an enemy of the state. In the collect second society we don't have to go around worrying about that kind of thing, because, you know, we live in a free society.
