.png){kind=icon}
Return
Zero-shot governance
.png){kind=icon}
OA
AI
DOI：10.1080/02680939.2026.2730191.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This article interrogates the concept of general-purpose AI governance. General-purpose AI governance – or ‘zero-shot governance’ – refers to a scenario in which instances of domain-agnostic generative AI intervene in policy decisions. This scenario is premised on the potential of the latest generation of AI foundation models to adapt to unfamiliar situations, requiring only relatively agile forms of customisation based on the introduction of domain-specific data. The central argument develops from the infrastructural analysis of Redbox, a now discontinued prototype designed to assist the daily cognitive work of civil servants in the UK. This tool is notable for representing a proof of concept for the integration of off-the-shelf ‘Large Language Models’ (LLMs) in the professional toolkit of policy. This integration is examined through intersecting technical, political, and cultural lenses. A review of the original codebase is followed by the analysis of the political and economic conditions in which Redbox and some its successors emerged. The overarching point emerging from the analysis is that the general-purpose nature of LLMs remains a structural component of the technology that can be mitigated but never ruled out. The conclusion reflects critically on the implications for education policy.
Keywords:
Generative AI
governance
zero-shot
Journal
IF:
3
Papers:
136
Citations:
3.1K
Organization
Cited Papers
No cited papers available