Morzsák

Oldal címe

Human-in-the-loop and machine-in-the-loop decision oversight in LLM-supported security compliance analysis

Címlapos tartalom

AI-supported security compliance analysis is helpful to process large volumes of security evidence, including enterprise documentation, cloud security posture management findings, configuration outputs, and control-related artifacts. However, while large language models and related AI techniques can improve speed and analytical coverage, they also introduce risks related to unsupported conclusions, unstable reasoning, weak traceability, and limited operational control. This paper reviews decision oversight model for LLM-supported security compliance analysis based on the combined roles of Human-in-the-loop and Machine-in-the-loop components. Human-in-the-loop oversight is treated as an explicit expert review. Machine-in-the-Loop oversight is treated as a structured set of automated control functions, including rule-based validation, source verification, confidence checks, contradiction detection, and escalation triggers. The paper reviews reliable enterprise adoption of LLM reasoning with deterministic controls and the need of accountable expert intervention. The proposed model is intended to support security compliance workflows where evidence sources are diverse and decision quality must remain auditable and reproducible.