Julie Cohen

Both in the U.S. and in Europe, initiatives for AI governance have focused principally on identifying and mitigating the risks created by AI models and their downstream uses rather than on those created by the datasets on which the models are trained. As this paper will explain, some of the most intractable dysfunctions of generative AI models involve datasets and data infrastructures. Effective AI governance requires an infrastructural turn in thinking about data and, along with it, a revised conception of public utility regulation that attends to matters of infrastructural configuration. The very large datasets amassed by dominant providers of AI-related services are rapidly taking on infrastructural characteristics and importance. First, the paper explains the significance of the infrastructure lens and sketches some of the distinctive implications of data infrastructures for governance of networked digital processes and the social and economic activities that they facilitate.  Next, it explores two interrelated problems manifesting within generative AI systems—simulation and sociopathy—that illustrate the extent to which the project of AI governance is, unavoidably, a data governance project. In brief, generative AI models trained on content from the public internet are also trained on data infrastructures that have been developed in particular ways for particular purposes and that encourage the production and spread of particular kinds of content. Last, it considers possible lessons to be learned from the growing movement to reinvigorate the concept of public utility within the law of regulated industries. Within that movement, there is definitional ambiguity about what a “public utility” is, what considerations a regime of public utility regulation ought to encompass, and what conduct it ought to require. Greater clarity on all of these matters, in turn, may aid policymakers both in avoiding likely regulatory dead ends and in identifying ways forward for development of a “public utility 2.0” construct for the AI era.