AI helps proactive disclosure of previously classified records

The State Department declassifies diplomatic cables and is the first-ever proactive disclosure made possible by machine learning and artificial intelligence.

A directive has been sent to U.S. embassies in Pakistan and India asking for an urgent assessment of both nations' financial and economic weaknesses.

A report outlining the strife inside the Bulgarian Socialist Party from the embassy in Sofia; an internal report on Secretary of State Madeleine Albright's visit with President Nelson Mandela in South Africa, prepared by the American embassy in Pretoria, was made public.

Those communications are among the several recently made public diplomatic cables from late 1997. Using a machine learning method that the State Department has been working on for the past year, the cables were declassified. Although no Freedom of Information Act requests were made for the cables, state authorities decided copies of the records could be made available to the public under the FOIA's "proactive disclosure" clause.

The Office of Global Information Services' deputy assistant secretary, Eric Stein, referred to it as the first-ever proactive disclosure of previously classified records made possible by machine learning and artificial intelligence

The State Department used years' worth of declassification judgments to build the machine learning algorithm. In deciding whether a record should be declassified, the algorithm is currently 97% as accurate as people, according to Stein, who spoke at a Digital Government Institute event on October 5.

According to Stein, the declassification pilot relies on human review at every stage to guarantee that the technology ultimately comes to the correct conclusion.