NSA surveillance program reaches 'into the past' to retrieve, replay phone callsThe National Security Agency has built a surveillance system capable of recording "100 percent" of a foreign country's telephone calls, enabling the agency to rewind and review conversations as long as a month after they take place, according to people with direct knowledge of the effort and documents supplied by former contractor Edward Snowden.
A senior manager for the program compares it to a time machine -- one that can replay the voices from any call without requiring that a person be identified in advance for surveillance.
The voice interception program, called MYSTIC, began in 2009. Its RETRO tool, short for "retrospective retrieval," and related projects reached full capacity against the first target nation in 2011. Planning documents two years later anticipated similar operations elsewhere.
In the initial deployment, collection systems are recording "every single" conversation nationwide, storing billions of them in a 30-day rolling buffer that clears the oldest calls as new ones arrive, according to a classified summary.
The call buffer opens a door "into the past," the summary says, enabling users to "retrieve audio of interest that was not tasked at the time of the original call." Analysts listen to only a fraction of 1 percent of the calls, but the absolute numbers are high. Each month, they send millions of voice clippings, or "cuts," for processing and long-term storage.
At the request of U.S. officials, The Washington Post is withholding details that could be used to identify the country where the system is being employed or other countries where its use was envisioned.
No other NSA program disclosed to date has swallowed a nation's telephone network whole. Outside experts have sometimes described that prospect as disquieting but remote, with notable implications for a growing debate over the NSA's practice of "bulk collection" abroad.
To be fair, their claim of "100% coverage" would be a lot less impressive if they revealed that the country in question is Tuvalu.
The part that impressed me was the mental leap to "why not just save all of it?"
The assumption by Schneier and the rest of us at this point is that all US calls are automatically transcribed to text files. Much more compressible, and all the best voice to text software work in the world has been for English. The only reason to record voice is if it's a language you can't machine transcribe at high fidelity.
Why not save all of it?
That seems to have been the thinking in a lot of the programs we're finding out about:
'Hey guys, if we tap this fibre we can get access to the between datacentre comms for Google!'
'Can't we just ask them for access, get a warrant even?'
'Yeah, but we might as well tap their fibre anyway, never know when it might come in useful'
Hey, guys, if we get the other two branches to declare that this Amendment doesn't say what it says, then the journalists who depend on us for their livelihood might still say that the government is constitutional! That would certainly come in useful.