In his renowned paper Do Artifacts Have Politics?, Langdon Winner (Daedalus, 109(1), 121–136, 1980) ascribes two political qualities to technological artifacts. Firstly, Winner shows that technological artifacts can be used to settle social or political problems, and can thereby structure social relations, reinforce vested interests, and engineer human relations (i.e., politicized artifacts). Winner also discusses a second political quality of artifacts, arguing that some technological artifacts practically require or strongly attract a specific set of social and political conditions to operate (i.e., inherently political artifacts). We apply the insights from Winner’s theory on the politics of artifacts to the use of artificial intelligence (AI) within government and argue that Winner’s thought-provoking analysis of inherently political artifacts deserves more attention. Analysing how AI is and can be politicized helps us to see what social or political problems AI can assist governments in solving and how it may do so with important public values in mind (e.g., value-sensitive design). Dissecting the inherent politics of AI enriches our understanding of how AI’s need for data, expertise, and infrastructure, requires a new, horizontally-orientated governance structure, while also attracting public-private partnerships and pre-sorting certain environmental and labor conditions. Recognizing the inherently political character of AI is necessary to be able to resist the increasing power of the private sector in government. This article provides new insights into how the socio-political arrangements that artifacts attract may threaten key public values and entrench existing power inequalities.