Guarding AI: defence strategies against emerging threats

Recent research from TrendAI, a business unit of Trend Micro, identifies AI skills as a potential target for cyber attackers, particularly in sectors such as security operations centres, financial services, healthcare, industrial systems and the public sector.

As organisations expand the use of AI automation to address skills shortages and operational demands, AI skills are increasingly encoding sensitive knowledge, decision processes and response workflows. If exposed, these skills could provide insight into how organisations detect threats, make decisions and respond to incidents, potentially giving attackers an advantage.

The study notes that security operations centres (SOCs) may be particularly exposed. Access to AI skills could reveal alert triage logic, correlation rules and response playbooks, which could in turn enable attackers to suppress alerts, evade detection or manipulate severity ratings. Similar risks are identified in other sectors, including trading threshold manipulation in financial services and interference with clinical decision-making in healthcare.

The report states that many traditional security tools are not designed to address this type of risk. AI skills are typically composed of unstructured text, requiring semantic analysis rather than signature-based detection. The public availability of AI skills in open repositories may further increase exposure.

The report outlines a kill-chain model describing how AI skills could be compromised and set out detection and governance recommendations. They advise organisations to treat AI skills as sensitive intellectual property and to apply appropriate access controls, monitoring and validation from the outset.

The findings suggest that organisations may need to review their AI governance and security practices to ensure that AI-related assets are appropriately protected and managed.