In recent years, a discourse of ‘ethical artificial intelligence’ has emerged and gained international traction in response to widely publicised AI failures. In Australia, the discourse around ethical AI does not accord with the reality of AI deployment in the public sector. Drawing on institutional ethnographic approaches, this paper describes the misalignments between how technology is described in government documentation, and how it is deployed in social service delivery. We argue that the propagation of ethical principles legitimates established new public management strategies, and pre-empts questions regarding the efficacy of AI development; instead positioning implementation as inevitable and, provided an ethical framework is adopted, laudable. The ethical AI discourse acknowledges, and ostensibly seeks to move past, widely reported administrative failures involving new technologies. In actuality, this discourse works to make AI implementation a reality, ethical or not.

近年来，由于广泛传播的AI故障，一种“道德人工智能”的讨论出现并获得了国际关注。在澳大利亚，有关道德AI的论述与公共部门AI部署的现实不符。本文利用制度上的人种学方法，描述了政府文档中描述的技术与社会服务提供中的技术之间的错位。我们认为，道德原则的传播建立了新的公共管理策略，并避免了有关AI开发功效的问题。取而代之的是将实施定位为不可避免的，并且在采用道德框架的情况下值得称赞。道德的AI话语承认并且表面上试图超越过去，广泛报道了涉及新技术的管理失败。实际上，这种话语使AI实现成为现实，无论是否符合道德。