Narratives of Power: AI, Agency and the EU AI Act

Public debate around AI has been characterised by hegemonic narratives that delineate the scope of the conversation, suppressing different voices and limiting nuanced discussion of the technology’s harms and benefits. Scholars have identified technological or enchanted determinism, the arms race mentality and the utopia versus dystopia debate as key narratives around AI. These ideas saturate the conversation, driven primarily by the tech industry, which positions itself as the (only) expert voice. The industry’s capacity to drive these narratives is facilitated by its established institutional authority and dominant presence in public discourse, as well as the knowledge asymmetry inherent in the proprietary, ‘black-box’ nature of AI systems. Tech firms leverage the technical opacity of AI and their powerful societal position to define the contours of public debate, assigning agency to AI itself and the tech industry; constructing users and civil society as passive observers or recipients; and establishing a frame for debate that focuses on extreme consequences or hyperbolic benefits. This study, conducted early in 2024, investigates the effects of this interpretive frame on the policy choices and communications of the EU AI Act, through a mixed-methods computational and critical discourse analysis. Within EU governing body texts, the same discursive patterns that map agency and erase the middle ground of debate in AI are found, helping to marginalise alternative perspectives from other stakeholders and prevent nuanced understandings of potential harms. In the context of the Act itself, these patterns reverberate in the assignment of agentic roles and the scope of the regulation.