Nowadays emotion AI is emerging as a forum of outrageously disruptive technology in the industry of artificial intelligence with consequences to distant as mental health diagnosis, teaching interfaces, testing vehicles, customer-focused and even human–computer interfaces. Going further, it is a technology encompassed in a web of technical obstacles and ethical and social measures. In this chapter, the problems associated with the modelling techniques include difficulty in modelling generalization, biases in datasets, cross-cultural variation by individuals on the expression of emotions in real time, performance restrictions and ability to capture subtle contextual differences of emotions. Besides, numerous ethical problems, including invasion of privacy, emotional surveillance, right to consents and manipulability of feelings and secrets during decision making processes etc., are critically analyzed proving that attempts to create interpretable and transparent AI systems become inevitable. To address these multi-dimensional issues, the chapter explores the future-forward solutions such as federated learning systems, edge computing systems coupled with privacy-sensitive AI applications, emotion-aware multimodal fusion models to meet individual user/ inter-user needs and culturally adaptive emotion-based models. It points to the necessity to have interdisciplinary synergy which is a combination of views based on the fields of psychology, computer science, ethics and policy as a guide to responsible use of Emotion AI all over the world. This chapter attempts to chart an architectural path on how to create Emotion AI systems that would be fair, transparent, circumstantial, as well as human-centric throughout the global paradigm by outlining current traps, and future various innovations.