This article is based on a roundtable held at the Association of Internet Researchers (AoIR) annual conference in 2017, in Tartu, Estonia. The roundtable was organized by the Oxford Internet Institute's (OII) Digital Ethics Lab (DELab). It was entitled "Artificial Intelligence and the Good Society". It brought together four scholars—Michael Zimmer, Stine Lomborg, Ben Zevenbergen, and Corinne Cath—to discuss the promises and perils of artificial intelligence (AI), in particular what ethical frameworks are needed to guide AI's rapid development and increased use in societies. The paper covers three case studies. They give a distinct overview of the ethical issues raised by the use of AI at different levels of analysis: top-down application of AI, bottom-up use of AI, and how academics and governments have reacted to these new challenges. From the case studies, four areas emerged. They represent some of the most topical ethical questions related to AI: (1) its uses, (2) its users, (3) its designers, and (4) the data that fuel it. Each of them provided a specific subset of ethical concerns that need further investigation. In conclusion, three recommendations are formulated for researchers and regulators to ensure the AI has a net-positive impact on society.