From self-checkouts to self-driving cars to social media newsfeed to surveillance systems, artificial intelligence (AI) is part of our lives whether we know it or not. Recent scholarship is attempting to explore how different places and people are involved in the development of AI. This chapter builds on this scholarship to extend the debate around the hidden and opaque networks of sub-contracting practices for AI training. In particular, it examines the ways in which lead firms develop their AI systems by tapping into a complex network of suppliers, platforms and human workers doing data work in some of the most marginalised parts of the world. The chapter shows that despite their touted nature as autonomous machines (i.e. without human control), AI is dependent on a global army of human labour performing a wide variety of data enrichment tasks. It then concludes by discussing implications of this data work on workers and places where AI gets trained.