• Authentic Leadership Enhances Safety Performance. • Knowledge Hoarding and Perceived Organizational Support as Mediators. • Work Passion Plays a Dual Role. • Novel Contribution to Leadership and Safety Literature in an emerging Arab country context. This study examines the impact of Authentic Leadership (AL) on employee safety performance, focusing on Safety Compliance (SC) and Safety Participation (SP). It also investigates the indirect effects of Knowledge Hiding (KHo), Perceived Organizational Support (POS), and work passion on the relationship between AL and employee safety performance. A self-administered questionnaire was distributed to randomly selected employees working in Abu Dhabi’s public sector organizations within hazardous industries in the UAE. The results revealed that AL positively influences both SC and SP. Additionally, KHo and POS were identified as partial mediators in the relationships between AL and SC, and AL and SP. Harmonious Work Passion (HWP) positively moderated the AL–SC relationship, whereas Obsessive Work Passion (OWP) negatively moderated both the AL–SC and AL–SP relationships. This research contributes to the relatively understudied literature on AL by exploring KHo, POS, and work passion as novel outcomes. To the best of our knowledge, it is the first study to examine the roles of KHo and POS in the link between AL and employee safety performance. The findings offer valuable insights for leaders in the UAE’s hazardous industries on leveraging AL to enhance employee safety performance.

• Evaluated ChatGPT and Gemini in stand-alone STPA and FRAM safety assessments. • The study used LH2 aircraft refuelling as a novel, complex case study. • ChatGPT outperformed Gemini in structured STPA analysis tasks. • Both LLMs struggled with systems thinking and output consistency. • Findings offer lessons for AI integration in system safety assessments. The advancement of large language models (LLMs) shows immense promise in many domains. However, their reliability is still questionable. This study aims to comparatively examine the performance of ChatGPT and Gemini in conducting a stand-alone systems-based risk assessment using System-Theoretic Process Analysis (STPA) and the Functional Resonance Analysis Method (FRAM). Our findings revealed that both LLMs demonstrated weaknesses in their analyses, with ChatGPT generally outperforming Gemini regarding response comprehensiveness and adhering to the prompted format. Specifically, LLMs failed to use systems thinking in their stand-alone applications and failed to follow up on previous prompt outputs. While LLMs can provide substantial amounts of information quickly, the effectiveness of LLMs in system safety assessment is contingent on addressing their limitations and implementing strategies to improve their capabilities.