With the rapid development of artificial intelligence, Security Large Models have emerged as a key technological means to enhance the security operations of hospital networks. While these models empower critical functions such as threat detection and automated response, they also introduce systemic ethical risks, including data privacy breaches, algorithmic bias, ambiguous attribution of responsibility, and social equity concerns. Based on deployment experiences in large tertiary hospitals and guided by both domestic and international ethical frameworks for medical AI, this study systematically identifies four core dimensions of ethical risks in hospital security models: data privacy, algorithmic trustworthiness, stakeholder accountability, and social equity. In response, we propose a four-pillar collaborative governance mechanism that integrates data, algorithm, application, and legal perspectives. Additionally, an algorithmic ethical impact assessment matrix is developed based on different hospital zones, aiming to foster security large models that are controllable, explainable, accountable, and socially inclusive within medical environments. This work provides theoretical and practical guidance for the ethical and standardized development of intelligent security operations in healthcare.