Autonomous vehicles (AVs) have developed quickly in the past few years and are poised to define the future of mobility. This can be attributed to tailwinds such as the increase in computational power, advances in the artificial intelligence (AI) field, and a favorable regulatory environment created by the impending climate crisis. From a sustainability point of view, AVs can be a major enabling technology for various Sustainable Development Goals (SDGs). Nonetheless, their system complexity and increased dependence on intractable AI algorithms should be addressed and studied, as they could inhibit some of the SDGs. In this work, a comprehensive description of the different sensors and AI algorithms used in AVs is given, as well as their governance and ethical implications. The development of complex deep learning (DL) algorithms is seen as a key factor in the development of more complex AVs, and in the use of camera-based AVs over LIDAR/radar based platforms. When it comes to decision-making, reinforcement learning (RL) is the key machine learning subclass allowing for an increase in the complexity of the actions being automated. Nonetheless, both DL and RL require vast amounts of training data, which should be heterogeneous and exhaustive enough to allow for the safe deployment of the AV into the real world. This poses clear issues related to privacy and data governance, which should be addressed both by institutions and AV manufacturers. Lastly, the increased complexity of AVs raises ethical and accountability issues, and the lack of international safety standards, such as those existing for normal vehicles, is seen as a major issue to be tackled.