Since at least the 1960s, scholars, pundits and the general public have been worried about AI hype – the tendency to project more power, agency and intelligence on AI than it really has. To counter this tendency, academics often warn against anthropomorphizing AI or assuming that its impacts on society are inevitable. While these are sensible cautions, taken too far, they risk ignoring the way the design of AI actually can shape and condition how power and authority is distributed in society – and in academia. How, then, to see AI? Is it merely a tool people use for their own ends? Or can it hijack these ends? This paper argues that Langdon Winner provides provocative frameworks for tackling these questions.