The Pentagon is on the hunt for an advanced artificial intelligence system designed for swift threat detection. This innovative engine will integrate various sensor data to create a unified intelligence overview. It aims to process and deliver alerts in mere seconds, enabling rapid response to threats. Crucially, human oversight will play a vital role in monitoring progress and making informed decisions to adapt to evolving threats.
A ground-based interceptor missile launches during a test in California
The Pentagon is seeking an artificial intelligence system capable of identifying and assessing space and missile threats within seconds, aiming to turn fragmented sensor data into actionable military intelligence before critical response time runs out. The US defence Innovation Unit (DIU) has invited proposals for a “Space Threat Intelligence Synthesis Engine” that would combine live video, satellite imagery, radar and sensor feeds, geospatial data and classified intelligence reports into a continuously updated threat picture. The proposed system is expected to process incoming information and produce “actionable, confidence-scored alerts” that can support human analysts, human-supervised operations or automated workflows. According to the DIU solicitation, existing systems “struggle to distinguish closely spaced objects, track emerging threats, and keep threat models current.” The Pentagon wants AI software that can overcome these limitations and provide a “continuously updated understanding of space and missile threats fast enough to drive operational decisions.” DIU wants the system to process data and display results with a latency of no more than five seconds, with two seconds as the preferred objective. It should handle regular data streams of around 20 to 30 megabytes per minute and bursts of up to 5 gigabytes per minute.
Why every second matters
Missile defence involves several stages, including detection, identification, tracking, threat assessment, decision-making and interceptor launch. Information from multiple sensors must be processed and correlated before an incoming threat can be assessed and a response initiated. The Pentagon's push for faster AI processing reflects the growing volume and complexity of data generated by missile and drone attacks. Human operators can face significant challenges when they have to rapidly analyse information from radar, satellites, video and other intelligence sources. AI could shorten this decision cycle by processing multiple feeds simultaneously, correlating information and updating threat assessments in real time. Instead of waiting for analysts to manually piece together a threat picture, the system could generate confidence-scored alerts and highlight emerging threats within seconds. The Pentagon's preferred two-second processing target could potentially provide commanders and defensive systems with additional time to assess an incoming threat and respond.
Human oversight remains central
The proposed system is not designed to eliminate human involvement. DIU says the technology should support “analyst-in-the-loop, human-on-the-loop, or automated workflows”, allowing the level of human involvement to vary according to operational requirements. The system would provide operators with human-readable threat visualisations and confidence-scored alerts, while AI would handle the rapid processing and correlation of large volumes of data. The broader objective is to reduce the time analysts spend assembling and interpreting fragmented information, while retaining human oversight over monitoring and decision-making in critical military operations.
