Sovereign AI
Sovereign AI is the emerging framework that asks whether national or community sovereignty has any meaningful application to AI systems trained on globally scraped data, operated on infrastructure concentrated in a handful of jurisdictions, and shaped by values reflecting the priorities of their developers. The concept emerged from the recognition that contemporary AI systems are not neutral tools but specific cultural and political artifacts — embodying particular linguistic priorities, particular knowledge bases, particular value commitments, and particular understandings of what intelligence is for. A nation that depends entirely on AI systems developed elsewhere has, in this framework, surrendered a form of sovereignty more consequential than territorial control. Chang's framework treats sovereign AI as a contemporary expression of the developmental sovereignty that successful developers historically asserted — the right to shape the technological infrastructure of one's own society according to one's own priorities, against the pressure to accept whatever the leading powers provide.
In the AI Story
The sovereign AI framework has gained traction across diverse political contexts. India's policy discourse has explicitly invoked sovereign AI as justification for domestic capability building. The European Union's AI strategy emphasizes European sovereignty over AI infrastructure. Multiple Global South nations have raised sovereignty concerns in international AI governance discussions. The shared element across these contexts is the recognition that adoption of foreign AI systems carries political and cultural consequences beyond the technical functionality the systems provide.
The challenge for sovereign AI is the gap between aspiration and capability. Asserting the right to shape AI development is not the same as having the resources to actually shape it. Building genuine domestic AI capability requires the kind of sustained, large-scale investment that Chang's developmental state framework specifies — and the international policy environment continues to constrain the toolkit available to developing nations attempting to mount such investment.
The relationship between sovereign AI and the broader Chang framework is direct. Sovereign AI is what infant industry protection looks like when the industry in question is artificial intelligence. The argument for sovereign AI is structurally identical to the argument for protecting nascent textile manufacturing in 1830s Germany or nascent semiconductor manufacturing in 1980s Korea — strategic state intervention to build domestic capability against established foreign competition, with the long-term goal of participating in the global economy as producer rather than merely consumer.
The opposition to sovereign AI follows the predictable pattern. Foreign AI providers argue that sovereignty requirements are 'distortionary' and 'fragmenting'. International institutions echo the argument with technocratic vocabulary. The position assumes that the current distribution of AI capability is natural and that interventions to reshape it are illegitimate — exactly the assumption that Chang's historical work demonstrates to be incorrect.
Origin
The phrase 'sovereign AI' came into common policy use around 2022–2023, driven by national AI strategy documents in India, the EU, and several other jurisdictions. The intellectual lineage includes earlier debates about digital sovereignty, data sovereignty, and technological sovereignty — concepts that have been articulated since the early 2000s as nations have grappled with the implications of dependence on foreign-controlled digital infrastructure.
The framework gained particular urgency after the 2022 release of GPT-3.5 and the subsequent recognition that frontier AI systems would have cultural and political consequences extending far beyond their technical capabilities. The recognition that a small number of American (and increasingly Chinese) companies would be making decisions affecting billions of people globally, with no accountability to those people, intensified the search for institutional frameworks that could reassert democratic and national agency over AI development.
Key Ideas
Substantive sovereignty. The recognition that AI dependence carries political and cultural consequences beyond the technical functionality, and that genuine sovereignty requires capability to shape AI development.
Capability gap. The distance between asserting sovereignty rights and possessing the resources to exercise them — a gap that requires the kind of sustained investment the developmental state framework specifies.
Infant industry parallel. The structural identity between contemporary sovereign AI advocacy and historical infant industry protection — strategic intervention to build domestic capability against established foreign competition.
Policy space contestation. The ongoing struggle over whether developing nations will retain the policy tools required to mount effective sovereign AI strategies, or whether the contemporary international order will lock them out of the toolkit.
Appears in the Orange Pill Cycle
Further reading
- Anu Bradford, Digital Empires: The Global Battle to Regulate Technology (Oxford University Press, 2023).
- Ulises Mejias and Nick Couldry, Data Grab: The New Colonialism of Big Tech and How to Fight Back (W.H. Allen, 2024).
- Government of India, 'IndiaAI Mission' policy documents (2024).
- European Commission, 'A European Approach to Artificial Intelligence' (2021).