Training military artificial intelligence systems requires massive datasets, but real-world combat data is scarce, classified, and often incomplete. Developers are now generating those scenarios synthetically.
Pennsylvania-based Cignal Defense is expanding its platform for training machine vision systems using simulated environments rather than operational data.
Its platform, Cignal Engine, produces physics-based synthetic data designed to help AI detect and interpret objects, threats, and anomalies across defense sensor systems.
Programs such as the Pentagon’s Project Maven rely on large datasets from drone, radar, and satellite sensors — data that can be difficult to access and rarely captures edge-case threats.
“The AI systems protecting this country, here and abroad, need a training ground built in America, for American mission requirements. Cignal Engine is that platform,” said Jaclyn Fiterman, CEO of Cignal Defense.
“Built in collaboration with DHS, it generates the sovereign synthetic environments that let AI teams tackle the hardest detection challenges in defense and security — drones, drugs, detonators, defects — without waiting for real-world incidents to define the threat.”
The technology was developed through the US Department of Homeland Security’s Silicon Valley Innovation Program and first applied to airport security screening systems.
Through the program, the company generated synthetic CT baggage scan data used to train AI models to detect weapons and explosives.
The project highlights how synthetic datasets are becoming critical for training AI systems to detect rare threats without relying on large volumes of real-world incident data.
Defense AI Trained in Simulation
The shift is spreading across the defense sector as militaries look to accelerate AI development while reducing reliance on classified or hard-to-obtain data.
The US Army is funding RF-Gen, a platform that generates synthetic radio-frequency datasets to train AI for electronic warfare scenarios.
Simulation platforms are also scaling across domains.
Luminary’s Physics AI Factory uses GPU-based environments to generate synthetic datasets for training AI used in drones and submarine design.
Meanwhile, BAE Systems is developing OneArc, a platform that integrates virtual environments, training systems, and AI to support mission rehearsal and operational testing.
