The US Department of Defense is expanding the use of generative AI to train security systems on synthetic threat data as part of a new contract with ZeroEyes Government Solutions.
Under the $1.19-million Phase II deal, ZeroEyes will develop AI-driven capabilities to generate high-fidelity threat models aimed at improving base-level detection.
These capabilities will be integrated into the company’s Automated Security Threat Recognition & Alerting (ASTRA) platform, which uses zero-shot techniques to detect emerging threats from limited data.
ZeroEyes’ existing systems focus on real-time detection of weapons, drones, and other threats across fixed and mobile sensors, providing the operational baseline for the expanded generative AI capabilities.
Early Efforts
The effort builds on earlier work funded through the Small Business Innovation Research (SBIR) program with the US Air Force.
During Phase I, ZeroEyes tested AI-based threat detection using indoor drones to improve situational awareness in confined spaces and close base security gaps.
As company chief executive Mike Lahiff noted at the time, AI-enabled platforms can deliver “actionable intelligence before a dangerous incident occurs,” supporting faster response and decision-making.
Shift Toward Synthetic Training Environments
The program is supported by the Pentagon’s Chief Digital and Artificial Intelligence Office (CDAO) and the US Air Force’s Digital Transformation Office (DTO), reflecting a broader move toward synthetic data in defense training.
The shift is already visible across defense programs, where synthetic data platforms generate physics-based environments to train AI on rare, operationally relevant scenarios.
