Der Spiegel and The Intercept have a new story about the NSA and Turkey based largely on NSA files Edward Snowden acquired. While the NSA is cooperating with Turkey's secret services and helps them to assassinate Kurdish separatists it is also intensely spying on the Turkish leadership.
That is all the way I would have expected.
But there is one detail in the story which, to my best knowledge, reveals a NSA capability that was so far only rumored about:
In January 2012, US officials proposed supporting Turkey in their fight against the PKK with diverse measures, including access to a state-of-the-art speech recognition system that enabled real-time analysis of intercepted conversations. The system can even search for keywords and identify the person speaking if a voice sample of that individual has been stored.
There was always some assumption that the NSA would store not only all the medadata of all phone calls but also the content. Barton Gellman had published about the MYSTIC and RETRO program which back in 2009 allowed phone call storage of all calls in Afghanistan for up to a month. Some people refuted that the NSA could or would do this in more countries or for a longer time arguing that the storage of sound files of the phone calls would require too much data space.
But when the NSA, as is now revealed, uses sophisticated automated speech-to-text systems then it has only to store the text data of phone calls which is at least one magnitude smaller in data size than the sound data.
Every NSA target, potentially ever human being, has now to assume that everything it says or hears on the phone will be recorded and automatically searched by keywords and then marked, categorized and stored forever by some NSA system.
This is, I believe, a whole new dimension of NSA spying that may well change the way people are used to communicate and the intensity in which they are willing to express themselves "in private". "In private" now hardly exists anymore.
