A recent study in the International Journal of Applied Decision Sciences explores how artificial intelligence (AI) can help the U.S. Army identify internal threats. This research focuses on the Army’s Insider Threat Hub, which evaluates risks from individuals reported for potentially dangerous behavior. They propose a deep learning tool designed to enhance the way these cases are prioritized and managed.

Internal threats differ significantly from external ones. Those with authorized access to critical data can cause serious damage, whether intentionally or accidentally. This risk can come from current or former employees and contractors. In military situations, data disruptions can be life-threatening.

The Army employs hundreds of thousands of personnel and receives countless threat alerts. Researchers note that there is no unified system to assess these reports, which makes it increasingly difficult to pinpoint actual risks, leading to a growing backlog of unresolved cases.

The study presents an innovative solution: a classification model developed from previously assessed cases that evaluates whether a person poses a low or high risk. This system enables staff to focus first on the most critical cases.

The model considers known factors, such as impulsiveness and aggression, along with situational signs like financial stress or personal trauma, to gauge an individual’s potential threat. The combination of these factors gives the most accurate insights.

Tests using the trained model on different historical data sets achieved a detection accuracy of 96%. The system was able to accurately assess the seriousness of most threats or, when incorrect, was slightly cautious by overestimating the risk—an outcome preferable to missing identifiable dangers.