NSA Eyes AI to Compress Weekslong Signal Triage

NSA cybersecurity directorate head David Imbordino said at the Billington Cybersecurity Summit in Washington this week that AI triage could compress the days or weeks analysts now spend sifting through intercepted signals into a much faster loop. The bottleneck, he argued on a panel with other Five Eyes officials, is volume rather than novelty.

"Volume has always been a problem," Imbordino said, but the agency is "looking at" how AI tools could let analysts "get to the nuggets" of disparate pieces of information much more quickly.

The directorate handles both intelligence collection and cyberdefense, ingesting foreign communications and electronic signals from radar and weapons systems. Triage here means sorting that inflow and flagging what a human should examine, not making analytic judgments.

Imbordino's remarks follow deputy director Tim Kosiba's late-August statement that NSA wants access to a wide range of commercial AI models. Both sit on the implementation track for a June executive order directing national security agencies to build classified tests for advanced AI hacking capabilities, with a voluntary 30-day pre-release access window for developers.

Imbordino's language is forward-looking ("looking at," "likely"), not a deployed system. Whether the tooling lands at scale depends on those model-access talks and on classified-capability thresholds still being defined.