AI, Digital Platforms, and the New Systemic Risk

Artificial intelligence (AI) is becoming increasingly embedded in digital, social, and institutional infrastructures, often merging with platforms to form hybrid structures. As a result, systemic risk has emerged as a critical but undertheorized challenge. In this paper, we develop a rigorous framework for understanding systemic risk in AI, platform, and hybrid system governance. We draw on insights from finance, complex system theory, and cybersecurity – domains where systemic risk has already shaped regulatory responses. We argue that recent legislation, in particular the EU’s AI Act and Digital Services Act (DSA), invokes systemic risk but relies on narrow or ambiguous characterizations of this notion. At times, the legislation reduces the source of risk to specific capabilities in frontier AI models; at other times, it attributes the risk to harms occurring in economic market settings. The DSA, we posit, actually does a better job at identifying systemic risk than the more recent AI Act. Our framework identifies systemic risks overlooked by the DSA and AI Act — including multi-agent interactions, discrimination at scale, and large-scale hallucinations — which can destabilize institutions despite falling outside current definitions. More generally, we present a framework for systemic risk, categorized into four levels: single-model, multi-model, model-platform, and model-institution integration. We then test the characterizations of systemic risk in DSA, the AI Act, and our own framework by applying them to three key yet contentious examples – large-scale discrimination, hallucinations, and environmental effects – to illustrate their respective strengths and limitations. Against this background, this paper proposes reforms that broaden systemic risk assessments, strengthen coordination between regulatory regimes, and explicitly incorporate collective harms. As localized AI and platform failures may increasingly escalate into structural disruptions, we provide a conceptual foundation and a policy-relevant diagnostic toolkit for governing the interplay of AI and platforms in complex, interconnected societies.