The National Security Agency is leaning into artificial intelligence tools to process the mass volumes of data it collects, allowing analysts to sort through tasks and find anomalies faster, a top official said Tuesday.
“I think right now we’re looking at how to use an AI where we could triage so much more,” David Imbordino, the head of the agency’s cybersecurity directorate, said on a panel with officials from the “Five Eyes” intelligence alliance at the Billington Cybersecurity Summit in Washington, D.C.
The agency ingests vast amounts of data every day, including intercepted foreign communications and electronic signals from radar and weapons systems. Imbordino’s directorate, in particular, focuses on both intelligence collection and cyberdefense.
“Volume has always been a problem,” he said, but AI has raised the prospect of getting “to the nuggets” faster than before. In the near future, analysts will likely use AI tools to look at disparate pieces of information and evaluate them “much more quickly” than current methods that can take days or weeks, Imbordino added.
His remarks highlight one of many ways the nation’s cyber and intelligence agencies are putting AI to work and show the omnipresent role such systems will play in day-to-day work in spy offices.
NSA has also taken a leading role in the Trump administration’s AI policy efforts as increasingly powerful tools promise to help defenders find and fix software flaws while giving hackers new ways to break into networks. Late last month, NSA deputy director Tim Kosiba said that the agency wants access to a wide range of models and is engaging heavily with frontier AI companies.
The agency is helping to implement a sweeping June executive order directing national security agencies to create classified tests for determining when AI systems have developed sufficiently advanced hacking capabilities to warrant additional scrutiny. The order calls, in part, for a voluntary arrangement through which developers can give the federal government access to qualifying models for as many as 30 days before releasing them to other trusted partners.
Read the full article here