Sustainable artificial intelligence (AI) systems are vital for ensuring reliable and ethical operations across diverse domains. However, these systems are vulnerable to threats that compromise their security, functionality, and sustainability. Adversarial attacks can exploit vulnerabilities in machine learning (ML) models. Nowadays, the role of threat modeling and attack simulation has become more prominent for information technology as well as operational technology. This chapter explores the role of threat modeling in sustainable AI systems, focusing on its applications in addressing adversarial risks, enhancing robustness, and ensuring ethical compliance. A comprehensive threat modeling framework is proposed to analyze the cyber threats in the ML-driven healthcare system for improving the security, sustainability, and resilience of the system. An industry standard tool, Microsoft Threat Modeling, is integrated into the proposed framework.