Their talking points seem to be sticking close to these two phrases: "We do not listen to your telephone calls," and "We do not target your emails."[1] Interpreted the way the intelligence community interprets these words, listening to telephone calls and targeting your emails probably does not include:
* listening to your Skype conversations
* storing your emails (it's already legal for them to access emails older than 180 days w/o a warrant) [1]
* auto-translating your emails and generating summaries [2]
Getting more cynical/paranoid:
* using these summaries and data to flag communications based on some 'suspicion' heuristic[3]
* auto-generate and robo-sign a subpoena
* emailing it to the FISA court
Given our known capabilities in these areas, and the intelligence and diligence of those in our intelligence community, it's not unlikely that they are automating a great deal of the analysis they conduct, using software that yields results by methods we'd have called artificial intelligence a decade a go.
I would love to know if it is legal for machine learning and intelligence to generate enough suspicion for a warrant. Given what happened in the finance industry with robo-signing, is it going to far to suspect that the same process might be automating the generation and approval of individualized warrants?
[1] Yesterday in an Obama in an interview said, "What I can say unequivocally is that if you are a U.S. person, the NSA cannot listen to your telephone calls, and the NSA cannot target your emails … and have not". Then seconds later repeated: "if you’re a U.S. person, then NSA is not listening to your phone calls and it’s not targeting your emails unless it’s getting an individualized court order". Why use the same words to be vague when different phrasing would have clarified?
Yep, you're right on it. This is exactly what I've been discussing in most of my comments over the last few days. You hear, just as in every other media-saturated issue, the same key phrases being repeated to absurdity, so as to both shape public opinion and debate and to divert attention away from asking more pointed questions and having to discuss the real reach of the programs.
* listening to your Skype conversations
* storing your emails (it's already legal for them to access emails older than 180 days w/o a warrant) [1]
* auto-translating your emails and generating summaries [2]
Getting more cynical/paranoid:
* using these summaries and data to flag communications based on some 'suspicion' heuristic[3]
* auto-generate and robo-sign a subpoena
* emailing it to the FISA court
Given our known capabilities in these areas, and the intelligence and diligence of those in our intelligence community, it's not unlikely that they are automating a great deal of the analysis they conduct, using software that yields results by methods we'd have called artificial intelligence a decade a go.
I would love to know if it is legal for machine learning and intelligence to generate enough suspicion for a warrant. Given what happened in the finance industry with robo-signing, is it going to far to suspect that the same process might be automating the generation and approval of individualized warrants?
[1] Yesterday in an Obama in an interview said, "What I can say unequivocally is that if you are a U.S. person, the NSA cannot listen to your telephone calls, and the NSA cannot target your emails … and have not". Then seconds later repeated: "if you’re a U.S. person, then NSA is not listening to your phone calls and it’s not targeting your emails unless it’s getting an individualized court order". Why use the same words to be vague when different phrasing would have clarified?
[2] 'FBI Documents Suggest Feds Read Emails Without a Warrant', ACLU: http://www.aclu.org/blog/national-security-technology-and-li...
[3] e.g., "John Smith and someone probably his wife had a conversation about a party with James, Linda, and Osama Bin Laden on October 1st"