How Human-Ai Decision Rights Shape Cybersecurity: Empirical Evidence On Data Breaches

Engineering notes

Engineering notes will be added by the aipentium editorial team.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为大语言模型、生成式AI、ChatGPT相关技术、计算机视觉、深度学习等高价值论文补充中文说明。

Original abstract

Artificial intelligence has become a strategic cornerstone of organizational transformation, offering the potential to enhance cybersecurity while introducing new risks. Drawing on Agency Theory, we develop a framework that distinguishes between Autonomous AI strategy and an Advisory AI strategy. Using a large language model –based classification of job postings linked to panel data on 3,006 U.S. firms, we empirically examine how these two AI strategies impact data breach risks. We find that an Autonomous AI strategy increases breach risk, whereas an Advisory AI strategy reduces it. Furthermore, greater AI deployment depth amplifies the effects of both strategies—intensifying risk under Autonomous AI while strengthening the protective benefits of Advisory AI. Moreover, stronger AI governance does not offset the risks of Autonomous AI but further enhances the security benefits of Advisory AI. These findings demonstrate that the security implications of AI depend critically on how decision authority is structured within organizations.

Links and sources

• PDF from original source

Comments