The paper seeks to discuss how smart home voice assistants contribute to inference-based privacy risk by passively gathering data from user voice interactions. While the most obvious risk of these systems is the storage of user commands, it also enables adversaries to make inferences about user data, from their daily routine to their possible health status, without having to breach any stored data. The paper extends previous works by adding new threat classes, discussing machine learning attacks in greater depth, and providing a forward-thinking analysis of possible solutions to these issues. Moreover, this paper will conclude with a special section providing authors with guidelines on how to verify their works' originality.