Associations of patient safety-related perceptions and lifelong learning-related attitudes with critical trust in AI/digital clinical tools among medical students: a cross-sectional study

Patient safety, lifelong learning, and the safe use of artificial intelligence (AI) are increasingly relevant to undergraduate medical education, but their interrelationships remain poorly understood. This study examined medical students’ patient safety-related perceptions, lifelong learning-related attitudes, and critical trust in AI/digital clinical tools. This single-center cross-sectional survey was conducted at Gaziantep University Faculty of Medicine, Türkiye, between April and May 2026. Of 366 submitted responses, seven from students younger than 18 years were excluded, leaving 359 participants. Study-developed Likert-type item sets assessed patient safety, lifelong learning, and approaches to AI/digital clinical tools. Exploratory factor analysis, informed by parallel analysis, was used to derive exploratory indices. Associations were examined using nonparametric tests and multivariable linear regression, with sensitivity analyses using complete item sets and excluding invariant respondents. Patient safety training during the previous 12 months was reported by 72 students (20.1%), and AI/digital health or clinical decision support system training by 50 (13.9%). Among clinical-phase students, 10 of 79 (12.7%) reported having encountered an adverse-event or near-miss reporting system. The exploratory patient-safety index was positively correlated with the exploratory critical-trust index (Spearman’s ρ = 0.453, p < 0.001). In multivariable analysis, the patient-safety index showed the largest standardized association with critical trust (B = 0.473, 95% CI 0.385 to 0.560; β = 0.507; p < 0.001). Clinical-phase status was positively associated with critical trust (B = 0.167, p = 0.010), whereas reported AI/digital health training was inversely associated (B = − 0.177, p = 0.018). The dichotomous any-use variable for digital clinical or AI-based tools was not independently associated with critical trust. In the complete-item sensitivity model, the patient-safety association persisted but was attenuated (β = 0.292, p < 0.001). Patient safety-related perceptions were associated with critical trust in AI/digital clinical tools, although the lower coefficient in the complete-item analysis indicates that the estimated magnitude was sensitive to how the scores were constructed; differences in item composition, measurement reliability, and wording-related structure may also have contributed. These exploratory cross-sectional findings do not establish causal or curricular effects but provide a rationale for longitudinal, interventional, and simulation-based studies examining how patient safety, lifelong learning, and critical evaluation of AI-supported recommendations may intersect in medical education.