Adapting Zero Trust to University Laboratory Safety Governance in the Artificial Intelligence Era

In the artificial intelligence (AI) era, the structure of safety hazards in university laboratories is changing, with unconventional hazards making up a growing share. As AI evolves from an auxiliary tool into an autonomous research collaborator, interdisciplinary experiments are becoming faster and more intense, producing hazards that are more dynamic, less predictable, and harder to assess and prevent. Most laboratory safety frameworks, however, still rely on static rules built for known risks, and these are poorly suited to such emerging threats. A shift from static rule-based control toward dynamic risk management is therefore needed. Adopting a conceptual mapping approach rather than an empirical study, this paper examines how Zero Trust—a cybersecurity framework based on continuous trust evaluation—could be adapted to university laboratory safety. It defines three adapted attributes (universal distrust with dynamic verification, least privilege with dynamic adjustment, and continuous monitoring with real-time early warning), identifies the main implementation barriers, and suggests targeted responses.