CTI-REALM is Microsoft’s open-source benchmark that evaluates AI agents on real-world detection engineering. It measures whether an agent can take cyber threat intelligence (CTI) and produce validated ...
In 2025, my team within the Soldier Evaluation Directorate won the U.S. Army Test and Evaluation Command (ATEC)’s AI Challenge with a tool that could ...
The Talon IQ testbed conducted combat air patrol and target engagement maneuvers controlled by ShieldAI's Hivemind AI, before switching back to Northrop Grumman's Prism AI. Northrop Grumman and Shield ...
Teams must also be able to review and interpret these results much faster to effectively guide engineering decisions within ...