AI security research misses five critical risks

×

A Lancaster University team has identified five major gaps in AI security research after a government-commissioned review for the UK’s Department for Science, Innovation and Technology (DSIT).

The researchers analyzed peer-reviewed AI security literature published since 2020, using data-driven methods to surface more than 9,100 sources, extract themes, and quantify where the field is still thin. Their conclusion: security research is not keeping pace with how AI systems are built, deployed, and retired.

The review highlights five areas that need more work:

• Data integrity: methods to formally guarantee or verify the safety, authenticity, and security of data used by AI systems, including training data
• Security of AI system infrastructure: stronger focus on the infrastructure behind generative and agentic AI systems, and what secure-by-design means for models such as large language models
• Securing against end-user risks: more study of how users can be harmed by or pose risks to AI systems, including through reliance on hallucinated or malicious outputs
• Data and model provenance: better understanding of how third-party AI models are developed, trained, and what security risks come with them
• Model disposal: safer ways to dispose of or decommission AI systems at the end of their lifespan

The authors also flag agentic AI systems as an emerging threat where security research remains especially limited, though they expect activity in that area to grow as such models become more common.

“The review findings highlight how the cybersecurity of AI goes beyond the models themselves, and that more research is needed to understand and secure these systems across their life cycle. With AI adoption across society increasing, this is crucial for the safe use of this technology.”

The review was led by Professor Nick Race, associate dean for research in Lancaster University’s Faculty of Science and Technology and professor of networked systems, with Dr. Tessa Wilkie also credited as an author.

Recommended reading

ServiceNow RCE flaw is now under active attack

“We were delighted to be invited by DSIT to lead a review into the AI security research landscape, supporting the government in its work to enhance the cyber resilience of the UK.”

The report, “Thematic review and gap analysis on AI security,” was published on gov.uk. The story was published on July 14, 2026.