by Ronan Farrow
June 28, 2013
Senators Mark Udall and Ron Wyden are upset about something, they just can’t say what. In a letter sent to the National Security Agency this week about a fact sheet on its surveillance programs, the senators complained about what they refer to only as “the inaccuracy”. The inaccuracy is “significant”. The inaccuracy could “decrease public confidence in the NSA’s openness and its commitment to protecting Americans’ constitutional rights”. But, because the information underlying it is classified, the inaccuracy can’t be described.
This is either a frustrating illustration of the absurdities of America’s secrecy regime, or the start of a pretty solid vaudeville act.
The frenzied public debate over the NSA leaks has focused on the correctness of the government surveillance programs themselves. But America cannot properly debate these and future surveillance efforts until it decides what can be debated.
As an official in the first Obama administration, I worked in jobs requiring top secret clearance. I know firsthand how essential secrecy can be to effecting policy goals and how devastating leaks can be. I navigated diplomatic relationships threatened by the indiscriminate release of WikiLeaks documents, and volunteered on the taskforce that sifted through them, piecing together the damage done. But it is also true that a culture of over-classification has shielded too much from public debate and that more could be disclosed without damaging the efficacy of intelligence programs.
Trillions of new pages of text are classified each year. More than 4.8 million people now have a security clearance, including low level contractors like Edward Snowden. A committee established by Congress, the Public Interest Declassification Board, warned in December that rampant over-classification is “imped[ing] informed government decisions and an informed public” and, worse, “enabl[ing] corruption and malfeasance”. In one instance it documented, a government agency was found to be classifying one petabyte of new data every 18 months, the equivalent of 20m filing cabinets filled with text.