The companies implicated in the slide deck have issued carefully-worded denials, claiming that they have never heard of anything called PRISM (likely, as that was probably an internal NSA codename not revealed to the outside world), have never provided the NSA with direct access to their servers (which could just mean that the NSA had to request items of data, or sets of items of data, and got an itemised bill for them).
Of course, this would mean that the NSA has had the task of wading through vast amounts of trivia: of social chatter, chain letters, forwarded amusing cat/sloth/lemur photos (which they'd have to check for steganographed terrorist plans, of course), mundane updates about people's lunch choices/music listening/reaction to last night's Game Of Thrones episode, online shopping receipts, steamy texts to lovers, drunkfaced party photos, viral ads, skinnerbox game invitations, complaints about traffic/public transport/coworkers and such. Though one wonders to what extent this can be automated. For decades, the US intelligence community has been investing millions in artificial intelligence research (a holy grail of CIA-funded research a while ago was the problem of “gisting”, or accurately summarising large amounts of text for human consumption; this is a hard problem, because it requires semantic knowledge about what the text is about). Meanwhile, in the private sector, data mining has shown uncannily accurate results, to the point where retailers have to insert a few deliberately inaccurate or useless coupons into the books they send to customers as not to freak them out with how much they know their true heart. (Remember the story about the angry father demanding why Target was sending his teenage daughter coupons for nappies and prams, and then apologising a few weeks later when she confessed that she was actually pregnant?)
If the NSA has had an firehose-like feed of personal information on millions of individuals for years, it's not unreasonable to expect that some proportion of the multi-trillion-dollar US “black budget” has been allocated to research into finding ways of aggregating, interpreting and processing this information to build up summaries or models of individuals. These could be automated dossiers with estimated personality profiles (“probabilities of paranoia: 23% issues with authority: 17%, narcissism: 27%, procrastination: 53%, adherence to routine: 61%. Most likely to fear: abandonment (41%), cancer (37%), rats (29%), exposure of peccadillos (23%). Probably responsive to: intimidation (43%), flattery (37%)”), which could be useful if the powers that be need to apply subtle, very precise pressure on a conveniently located bystander to use them against someone like al-Qaeda or Occupy. If they have real-time information, such as the mobile phone metadata (and, even omitting the content of conversations, having a record of the location of a person's phone can reveal a lot about what they're doing), they could even get alerts when somebody deviates from their routine more than they typically do; a dive into their private data would reveal whether they're planning a surprise anniversary party for their spouse or a terrorist attack. (Spoiler: it's almost never a terrorist attack.)
Of course, what the social and psychological effects of such surveillance are is another question. If there is a class of watchers, who can peer into the deepest secrets of the rest of the population, would their attitude to the pitiful, flawed wretches before them, with their pathetic little sins and failings, not be one of contempt? Would they not start regarding the rest of the population as little more than cattle, much as the participants in the Stanford Prison Experiment did?
Please keep comments on topic and to the point. Inappropriate comments may be deleted.
Note that markup is stripped from comments; URLs will be automatically converted into links.