AIGF-DLM: Cognitive-Behavioural Governance for GenAI Data Leakage

Abstract

AI governance frameworks for GenAI data leakage remain inadequate because they address technical and regulatory dimensions in isolation while neglecting the cognitive-behavioural mechanisms driving human data-sharing behaviour. This paper makes two contributions. First, the AI Governance Framework for Data Leakage Mitigation (AIGF-DLM) integrates five interdependent governance dimensions — Data Governance, Technical Controls, Organisational Governance, Regulatory Compliance, and Continuous Assurance  uniquely incorporating synthetic data leakage testing, agentic AI governance, and cognitive-behavioural risk factors. Second, the Cognitive Synthesis Risk Governance (CSRG) theoretical framework introduces three formal propositions linking cognitive risk culture, cognitive load, and moral sensitivity to governance effectiveness. Systematic secondary data analysis of 30 findings from 14 industry reports published by 13 independent organisations demonstrates convergent support for all three propositions. Comparative analysis confirms that the AIGF-DLM addresses critical gaps in six leading frameworks.