AI Security Review: Key Themes & Research Gaps Identified

A new thematic review and gap analysis on AI security has identified 12 key themes in research and five prevalent areas needing more study. Lancaster University conducted this review, commissioned by the Department for Science, Innovation and Technology. The review looked at peer-reviewed research on AI security from the last five years. It used data-driven techniques to map the research into themes. These themes include Data and Privacy Risks, Supply Chain Vulnerabilities, and Security Governance and Regulation. The study also found significant gaps. More research is needed on areas like verifying AI system data integrity and understanding how end-users can pose security risks. There's also a call for more focus on the cybersecurity of Agentic-AI systems. This kind of analysis helps ensure AI systems are secure as they become central to growth strategies.