AI-based threat hunting focuses on proactively investigating Linux system activity to identify hidden threats, suspicious behavior, and potential attack techniques that may not be detected by conventional security monitoring. It involves examining system events, analyzing kernel system-call patterns, identifying deviations from normal behavior, and developing investigation hypotheses to uncover potential malicious activities. By combining local large language model-assisted hypothesis generation with deep learning-based behavioral anomaly detection, this approach helps security analysts investigate possible privilege escalation attempts, identify unusual system interactions, and prioritize suspicious activities for further validation. The resulting insights support evidence-driven investigations, earlier identification of emerging threats, improved understanding of attacker behavior, and timely implementation of appropriate defensive measures.