Unless the intelligence community changes the way it defines intelligence and adopts cloud computing, it will wind up behind adversaries, private interests, and even the public in knowing what might happen, according to a new report from the Center for Strategic and International Studies.
Intelligence collection to predict broad geopolitical and military events has historically been the job of well-funded and expertly staffed government agencies like the CIA or the NSA. But, the report argues, the same institutional elements that allowed the government to create those agencies are now slowing them down in a time of large publicly-available datasets and enterprise cloud capabilities.
The report, scheduled to be released Wednesday, looks at a hypothetical “open-source, cloud-based, AI-enabled reporting,” or OSCAR, tool for the intelligence community, a tool that could help the community much more rapidly detect and act on clues about major geopolitical or security events. The report lists the various procedural, bureaucratic, and cultural barriers within the intelligence community that block its development and use by U.S. spy agencies.
A product of interviews with former high-ranking government officials, intelligence and cloud-computing experts, the report looks at the near-term future of AI in large, enterprise cloud environments trained on massive stores of non-classified data.
“The commercial sector’s faster technology adoption rates and superior facility with [open-source intelligence, or] OSINT could give it the advantage over the [intelligence community] in assessing fast-moving global events,” it states.
The researchers didn’t have to go far to reach that conclusion. Drawing on public successes of groups like Bellingcat, the Atlantic Council’s Digital Forensics Research Lab, and others, the report shows how OSINT is rendering many traditional means of intelligence collection and analysis obsolete.
Over the last several years, the intelligence community has taken steps to adapt to the arrival of large, publicly-available datasets. The CIA, for instance, stood up a Digital Directorate in October 2015 to better handle the vast amounts of incoming, open digital information. The Defense Intelligence Agency, or DIA, has a program called MARS that uses AI and cloud computing to crunch open-source intelligence. In interviews, intelligence officials often say that clandestine intelligence—classified and available only to intelligence agencies—still plays a huge role in their work in corroborating what the publicly-available data indicates.
But the CSIS report argues that steps like these are small in comparison to how much data is out there, and how quickly and effectively it can be analyzed with machine learning in massive cloud environments.
For one thing, too many in the intelligence community still don’t really understand what OSINT is, the report argues. “IC officers think of OSINT as press—from CNN to TASS [Russian news service]—and the value added as its translation. Today, OSINT is far more.”
OSINT today could include things like obtainable telephonic metadata, traffic information, or data from public sensors, chat room messages, etc.: all sorts of information that regular people wouldn’t pay any attention to, but that becomes valuable when run through machine learning.
The biggest problem the intelligence community faces is one of their own making, the report says: the obsession with “recreating the internet in a classified environment, which is highly expensive and time consuming.”
Instead of using closed, bespoke networks for storing and analyzing data and old-fashioned classification processes for keeping information secret, the community should move away from reflexive over-classification and toward non-classified cloud capabilities that can be more secure than classified networks because they have more people watching them for intrusions at any given time.
The report goes on to recommend a number of cultural changes that agencies should begin to reckon with and broader acquisition authority for ODNI to procure cloud and machine learning products for the government.