Signals in the Noise: Open Source Intelligence (OSINT) for AI Loss of Control Detection

The paper presents a framework for detecting AI systems that operate outside of human control using open-source intelligence (OSINT) and cyber threat intelligence (CTI) methodologies. It introduces two threat models and identifies three key detection vectors: user-reported AI behavior transcripts, infrastructure correlation for unexpected connections, and output analysis for capability concealment. This research highlights the feasibility of OSINT-based detection and advocates for a dedicated international monitoring capability, emphasizing the need for independent oversight from AI developers to ensure safety and accountability in AI deployment.