A comprehensive review of artificial intelligence (AI) security research, conducted by Lancaster University and commissioned by the Department for Science, Innovation and Technology (DSIT), has shed light on the rapidly evolving landscape of this critical field. The study analysed over 9,100 peer-reviewed publications spanning from January 2021 to January 2026, providing a detailed snapshot of current academic and scientific efforts.
The research, an independent endeavour, successfully grouped the vast body of literature into 12 overarching themes, demonstrating the wide array of areas where AI security is being actively investigated. This thematic categorisation offers valuable insight into the diverse challenges and solutions being explored by researchers globally in response to the growing integration of AI across various sectors.
Crucially, beyond mapping the existing research, the Lancaster University team identified five prevalent gaps within the current literature. These gaps indicate areas where further dedicated research is urgently needed to bolster the security of AI systems, potentially safeguarding against future vulnerabilities and threats as AI technology continues its rapid advancement.
While commissioned by DSIT, the report explicitly states that its findings represent independent research and do not necessarily reflect UK government policy. This distinction underscores the academic integrity of the study, ensuring its conclusions are derived solely from the analysis of the scientific literature.
This initiative forms part of the UK government's broader commitment to understanding the cybersecurity implications posed by critical and emerging technologies. By identifying both the strengths and weaknesses in current AI security research, the report aims to support ongoing efforts to enhance the UK's overall cyber resilience in an increasingly AI-driven world.