Voice-Activated Artificial Intelligence (VAI) is increasingly ubiquitous, whether appearing as context-specific conversational assistants or more personalised and generalised personal assistants such as Alexa or Siri. CSCW and other fields have regularly studied the (positive and negative) social consequences of VAI design and deployment. One particular focus has been questions of gender, and the implications that the (often-feminine) gendering of VAIs has for societal norms and user experiences. Studies into this have largely elided transgender (trans) existences; the few exceptions to this operate largely from an external and predetermined idea of trans and/or non-binary user needs, centered on representation. In this study, we undertook a series of qualitative interviews with trans and/or non-binary users of VAIs to explore their experiences and needs. Our results show that these needs are far more than simply improvements to representation, and that users raise substantial concerns around the underlying framing of gender by even well-intentioned developers, the privacy and safety implications of ubiquitous VAI, and the motivations of the vast for-profit companies that deploy much of this technology. We provide both immediate recommendations for designers and researchers seeking to create trans-inclusive VAIs, and wider, critical proposals for how we as researchers go about assessing technological systems and appropriate points of intervention.