Aim: To examine nurse managers’ ethical decisions and the concordance and response stability of generative artificial intelligence in simulated management scenarios. Background: Nurse managers balance obligations to patients, staff, and organizations. Evaluating both agreement and repeated response variability is necessary when considering artificial intelligence for ethical deliberation and organizational governance. Design: An exploratory study combining a cross-sectional census at a Spanish public university hospital with repeated evaluation of three models using identical sce-narios. Methods: Models received the complete questionnaire in new web conversations. Analyses examined human response distributions, management-level variation, model concordance, and within-model stability. Reporting followed the Strengthening the Reporting of Observational Studies in Epidemiology checklist. Results: Nurse managers converged most strongly on protective and fairness-related actions. Staffing and informed consent elicited more heterogeneous choices. Overall concordance was similar, although two models were highly stable and the third varied more across repetitions. Scenario-specific agreement differed. All models consistently diverged from the human mode in conscientious objection and informed consent. Exploratory comparisons did not establish model superiority or equivalence. Conclusion: Repeated evaluation provides a more informative account of model behavior than a single response. Aggregate agreement con-cealed distinct response patterns. Stability, professional concordance, and ethical defensibility represent separate questions; consistent outputs may still diverge from professional preferences. These findings do not establish clinical effectiveness or ethical correctness. Implications for Nursing: Leadership education should combine ethical reflection with critical appraisal of model outputs and context-sensitive justification of decisions. Implications for Health Policy: Organizational govern-ance should retain accountable nursing oversight, require local evaluation and review after model updates, and define procedures for resolving consequential disagreement.