Artificial Intelligence (AI) systems are reshaping users’ interactions with products and bringing new challenges for Human-Computer Interaction (HCI). Such a scenario highlights the need to apply a fundamental principle for human-centered design: transparency. However, “transparency” is still a blurry concept in the context of AI systems due to the wide availability of definitions and analog terms across the literature of varied fields. Therefore, this paper aims to outline the concept of transparency for AI systems through the perspective of the HCI field by highlighting differences and similarities across varied definitions. We showed that transparency might be considered both as a driver for users’ perceptions and other system characteristics or as an outcome of developmental techniques that aim to make systems more explainable. Whereas this work does not intend to provide a new definition for transparency, we expect our discussion to support further explorations on this concept.