Competing worldviews in AI governance: Directed content and critical discourse analysis of five official documents

Abstract

AI-governance documents are often taken to speak for the polities issuing them. Using directed content and critical discourse analysis, we examine five documents issued by two U.S. administrations, China, the EU, and BRICS, along 11 analytical dimensions, asking whether that assumption holds and where they diverge and converge. The assumption fails most clearly when a polity speaks twice: the two U.S. documents, from 2022 and 2025, take opposed positions, and China’s 2017 plan, still operative, asserts sovereignty on its own terms while the 2025 declaration China co-signed centers the United Nations. The U.S. pair shows rival discursive projects contesting the national interest within one state. No dimension yields American-European and Chinese-BRICS blocs: self-narrative comes closest, joining both U.S. documents to the EU’s but splitting China from BRICS. The Biden-era document matches the BRICS declaration on six of the 11 dimensions, and the Trump and EU plans, far apart on regulation, agree on the main risk they identify. Three dimensions, naming an enemy, open source, and environmental limits, are examined in depth as the fault lines dividing the documents. These documents carry competing worldviews of what AI is and the order it belongs in.