The concept of sovereign AI has gained traction among governments worldwide, promising nations the ability to control their own artificial intelligence systems without depending on foreign technology giants. Yet according to author and activist Cory Doctorow, writing for The Register, this idea amounts to little more than empty political rhetoric that distracts from genuine problems in how AI develops and spreads.
Doctorow argues that calls for sovereign AI often serve as cover for protectionist policies or as marketing slogans rather than serious technical proposals. When politicians talk about building homegrown AI to protect national interests, they rarely address the practical barriers that make true independence nearly impossible. Training modern large language models requires enormous computing resources, specialized hardware, and vast datasets that few countries can assemble on their own. Even those with the financial muscle to attempt it quickly discover that the underlying components remain concentrated in the hands of a small number of American and Chinese companies.
The piece highlights how nations pursuing sovereign AI initiatives frequently end up partnering with the very foreign firms they claim to be escaping. Saudi Arabia, France, and India have all announced ambitious programs to develop independent AI capabilities, yet their projects typically rely on chips from Nvidia, cloud infrastructure from Amazon or Microsoft, and foundational models adapted from OpenAI or Meta. This pattern suggests that sovereign AI functions more as a branding exercise than a path to actual autonomy.
Doctorow points out that the term itself borrows from legitimate discussions about data sovereignty and digital independence, but applies those ideas in ways that do not translate well to AI systems. Data sovereignty makes sense when governments want to prevent sensitive citizen information from being stored on foreign servers. AI sovereignty, by contrast, runs into fundamental problems because intelligence emerges from complex interactions between models, training data, and deployment environments that resist clean national boundaries.
The article examines several high-profile sovereign AI efforts and finds them lacking in substance. When the European Union talks about creating European AI champions, the reality often involves heavy subsidies for local startups that still depend on American foundational technology. China’s push for self-reliant AI has produced impressive results in some areas, but even there the country continues to import critical semiconductor manufacturing equipment and relies on global supply chains for rare earth minerals essential to hardware production.
One of the more compelling sections of Doctorow’s analysis focuses on the illusion of control. Proponents of sovereign AI suggest that owning the models will allow governments to align them with national values and security requirements. This assumption overlooks how AI systems trained on internet-scale data absorb patterns and biases that transcend any single country’s influence. Once a model has been exposed to the full spectrum of human knowledge available online, attempting to retrofit it with specifically French or Brazilian characteristics becomes an exercise in damage control rather than genuine shaping.
The piece also addresses the security implications that sovereign AI advocates frequently cite. The argument goes that depending on foreign AI leaves nations vulnerable to backdoors, hidden biases, or sudden service cutoffs during geopolitical tensions. While these concerns contain elements of truth, Doctorow counters that building domestic alternatives creates new vulnerabilities. A government-controlled AI system might be more susceptible to domestic political interference, surveillance overreach, or simple technical incompetence. The history of state-run technology projects offers numerous examples where national pride led to systems that were both expensive and ineffective.
Doctorow suggests that instead of chasing the fantasy of complete AI independence, countries should focus on practical measures that actually increase their technological resilience. These include investing in open source AI development, creating transparent auditing mechanisms for deployed systems, developing strong domestic computing infrastructure, and establishing international agreements that prevent any single company from dominating critical AI capabilities. Such approaches acknowledge the inherently global nature of modern AI research while still protecting legitimate national interests.
The article draws parallels with previous technological sovereignty movements that ultimately proved more symbolic than substantive. The push for national champions in semiconductors during the 1980s and 1990s produced mixed results at best. Many countries that tried to build their own chip industries from scratch ended up with expensive facilities producing outdated technology. The software industry followed a different pattern, with open source movements demonstrating that collaborative development across borders could create more robust alternatives than isolated national efforts.
Particularly interesting is Doctorow’s examination of how sovereign AI rhetoric serves different purposes in different contexts. In authoritarian states, it becomes a tool for justifying greater control over information flows and citizen behavior. In democratic nations, it often functions as industrial policy dressed up in patriotic language. In both cases, the language of sovereignty obscures more immediate questions about who actually benefits from these massive technology investments and whether they address real societal needs.
The piece notes that genuine advances in AI accessibility have come not from government declarations of independence but from researchers and developers who share their work openly. Projects like Hugging Face have democratized access to AI tools far more effectively than any national program. When smaller countries or independent organizations can fine-tune existing models for local languages and cultural contexts, they achieve a form of practical sovereignty without needing to replicate the enormous costs of training foundation models from scratch.
Doctorow warns that the sovereign AI narrative risks diverting resources from more pressing challenges. While politicians debate how to build national AI systems, basic questions about AI safety, labor displacement, environmental impact, and ethical deployment receive less attention. The computational resources required for frontier AI development consume electricity at scales that strain power grids and contribute significantly to carbon emissions. Countries serious about technological independence might better serve their citizens by addressing these concrete problems rather than pursuing the chimera of completely autonomous AI.
The analysis extends to the economic dimensions of AI development. The capital requirements for training state-of-the-art models have grown so extreme that even the largest technology companies struggle to fund successive generations of systems. This reality makes the idea of dozens of nations each maintaining their own frontier AI labs seem particularly unrealistic. Smaller countries would be better served by focusing on applications, implementation, and governance rather than competing in the raw compute arms race.
Throughout the article, Doctorow maintains that the problems with sovereign AI stem not from any hostility to national interests but from a clear-eyed assessment of technological reality. AI systems today represent the cumulative work of researchers across many countries over decades. Attempting to nationalize this shared intellectual commons creates more problems than it solves. The better approach involves finding ways for nations to contribute to and benefit from global AI development while maintaining appropriate safeguards and oversight.
The piece concludes by suggesting that policymakers should treat AI as a global infrastructure requiring coordinated governance rather than as territory to be claimed by individual nations. This perspective aligns with how societies have successfully managed other transnational technologies like aviation, telecommunications, and the internet itself. While complete international agreement remains elusive, practical cooperation on standards, safety protocols, and access to computational resources offers more promise than isolated sovereignty projects.
By framing sovereign AI as primarily a political slogan rather than a feasible technical goal, Doctorow’s argument encourages more honest discussions about what nations can and should expect from artificial intelligence. The technology will continue advancing through complex networks of collaboration, competition, and shared knowledge that no single country can fully control or opt out of. Recognizing this interconnected reality represents the first step toward developing policies that actually serve the public interest rather than feeding nationalistic fantasies about technological independence.
Countries that accept the global character of AI development can still make strategic choices about which capabilities to develop domestically, which partnerships to pursue, and what regulatory frameworks best protect their citizens. This pragmatic approach stands in stark contrast to the grand but ultimately hollow promises of complete AI sovereignty that continue to feature prominently in political speeches and policy documents around the world. The real work of building beneficial AI systems requires focus, cooperation, and technical honesty rather than slogans that obscure the genuine difficulties involved.
WebProNews is an iEntry Publication