Who really controls artificial intelligence? #70
Sovereignty is no longer territorial but cognitive: AI increasingly mediates how we think, decide, and understand the world. In this new world, power lies in shaping human perception and knowledge.
(Service Announcement)
This newsletter (which now has over 6,000 subscribers and many more readers, as it’s also published online) is free and entirely independent.
It has never accepted sponsors or advertisements, and is made in my spare time.
If you like it, you can contribute by forwarding it to anyone who might be interested, or promoting it on social media.
Many readers, whom I sincerely thank, have become supporters by making a donation.
Thank you so much for your support, now it's time to dive into the content!
Throughout history, power has been defined by control over strategic resources. The Renaissance printing press shaped minds through ideas, the industrial age measured sovereignty by territorial control and energy sources, and the twentieth century elevated nations commanding oil, nuclear capabilities, and manufacturing infrastructure. Today, we witness an entirely different paradigm that transcends traditional boundaries and challenges our fundamental understanding of sovereignty itself.
If sovereignty was once territorial, today it has become cognitive, residing in the ability to shape and mediate human thought through technological intermediaries. We are no longer discussing mere data sovereignty, but something far more profound: sovereignty over the very processes of reasoning, decision-making, and knowledge creation increasingly mediated by artificial intelligence systems. These technologies actively process, interpret, and generate knowledge in ways that fundamentally alter how we perceive reality and understand our world.
The architecture of cognitive power
Cognitive sovereignty operates through a complex chain of dependencies spanning technical, economic, and political dimensions. Unlike traditional forms of power visualized through territorial maps or hierarchies, this new architecture determines who can think what, who can know what, and ultimately who shapes our digital age’s cognitive infrastructure. At the foundation lies data ownership and infrastructure. Entities controlling vast data centers and cloud computing facilities hold the raw material from which AI systems derive their capabilities. These repositories of human knowledge, behavior, and expression constitute the training ground for machine learning models, and their concentration in relatively few corporations creates fundamental asymmetry in cognitive power, granting unprecedented insight into human thought patterns at previously unimaginable scales. When data flows predominantly through servers owned by particular corporations, those entities gain not merely economic advantage but epistemic privilege, the ability to observe, analyze, and understand human cognition at depths never before possible.
The second layer involves constructing foundational models themselves. Developing large language models, computer vision systems, and other AI architectures requires not only massive computational resources but also sophisticated expertise in machine learning, substantial financial capital, and access to cutting-edge research. Currently concentrated among technology giants primarily in the United States and China, along with well-funded startups, these models become cognitive engines powering countless applications and shaping how billions interact with information, make decisions, and form understanding. When a relatively small group of engineers and executives determines the architecture, training methodology, and capabilities of foundational models, they effectively architect our collective cognitive infrastructure, establishing the parameters within which digital reasoning and knowledge generation can occur.
Equally significant is the authority defining rules governing model behavior and alignment. Through policies on training data curation, content filtering, response alignment, and acceptable use parameters, model creators exercise editorial control over machine cognition boundaries. These decisions determine which knowledge an AI system can access and articulate, which perspectives it privileges or marginalizes, and which questions it answers or deflects. The implications extend far beyond technical considerations, touching fundamental questions of intellectual freedom, epistemic diversity, and justice in knowledge distribution. When proprietary models reflect particular cultural contexts or corporate interests, they risk encoding specific ideological frameworks as universal truths, potentially creating cognitive monocultures that stifle alternative ways of knowing and understanding.
Underpinning everything lies physical infrastructure: semiconductors and specialized hardware providing computational substrate for machine learning. The supply chain concentrated in Taiwan and South Korea reveals another sovereignty dimension increasingly subject to geopolitical competition. Access to advanced computing hardware has become strategic resource, with nations imposing export controls to maintain technological advantages. Control over semiconductor manufacturing determines which entities can participate meaningfully in advanced AI development.
Geopolitical dimensions of cognitive sovereignty
Three distinct visions have crystallized, revealing contrasting values for managing technology, society, and state power relationships. The American approach embodies faith in private sector innovation as the primary engine of progress. Major corporations operate with substantial autonomy, constrained primarily by market forces and relatively light-touch governance frameworks. This model has produced remarkable innovations and dominant platforms shaping global digital infrastructure, from search engines revolutionizing information access to social media reshaping human communication, from cloud computing enabling unprecedented computational scale to AI services transforming everything from healthcare to creative production. However, this concentration raises profound questions about accountability and transparency. When entities building cognitive infrastructure answer primarily to shareholders rather than citizens, with opaque decision-making behind proprietary secrecy, we must question whether such arrangements serve democratic governance and human flourishing. The American model offers dynamism and rapid innovation but struggles with fundamental questions of equity, broad access, and meaningful democratic oversight of increasingly powerful technologies.
China pursues centralized state sovereignty over AI development and deployment. The government articulates explicit national strategies, channeling substantial resources while maintaining tight regulatory control. This enables coordinated mobilization toward strategic objectives, potentially accelerating innovation and ensuring state alignment. However, integrating AI into social control and surveillance systems raises significant concerns about privacy, freedom, and autonomy. The Chinese model demonstrates cognitive sovereignty wielded as state power instrument, both domestically and through growing technological influence across developing nations.
The European Union charts a third path through regulatory frameworks and rights-based approaches. The AI Act represents the most comprehensive effort to govern AI according to transparency, accountability, and human rights principles. Yet Europe faces a fundamental challenge: can sovereignty be exercised through regulation alone when underlying infrastructure and foundational models remain controlled by foreign entities? The question of whether Europe can achieve meaningful cognitive autonomy through regulation, public investment, and alternative model support remains critically open.
From digital colonies to cognitive commons
For most individuals and organizations, the relationship with AI systems resembles colonial subjects dependent on infrastructure controlled by distant powers. We have become cognitive colonists using models we do not own, developed through processes we cannot inspect, trained on opaque data collections, governed by unilaterally changing terms. These systems learn from our interactions, extracting value from our cognitive contributions while offering little transparency into how that value is captured or deployed.
When our access to knowledge and decision-making capacity become mediated by AI systems we do not control, what does meaningful sovereignty over our cognitive lives mean? The answer cannot be individual withdrawal, as these systems are deeply embedded in modern life infrastructure, from the search engines mediating our access to information to the recommendation algorithms shaping our cultural consumption, from the automated systems managing critical infrastructure to the AI assistants increasingly integrated into professional and personal workflows. Instead, we must consider alternative models distributing cognitive sovereignty more equitably across society, creating new institutional arrangements that balance innovation with accountability, efficiency with transparency, and technological advancement with human agency.
Open-source and open-weight models making AI architecture and parameters publicly accessible offer promising directions. Projects like Mistral, Llama, and Falcon demonstrate viability of creating powerful systems outside proprietary models. These represent nascent cognitive resistance, building alternative infrastructures enabling transparency, customization, and community control. When model weights are public, researchers inspect behavior, developers adapt to specific contexts, and communities collectively govern evolution. This shifts from cognitive colonialism toward cognitive commons, shared resources governed for broader community benefit rather than narrow interests.
However, open-source alone cannot fully address sovereignty challenges. Sophisticated AI development still requires substantial computational resources, expertise, and coordinated effort. True cognitive sovereignty requires not only technology access but meaningful agency over its development and deployment. This suggests need for intermediate institutions, cooperative structures, and public investments aggregating resources while maintaining democratic accountability.
Toward distributed cognitive sovereignty
Distributed cognitive sovereignty offers a compelling framework recognizing both technical realities and political imperatives. Rather than recreating concentrated sovereignties at different scales, it imagines pluralistic ecosystems where multiple communities participate in creating, governing, and deploying intelligent systems according to diverse values and needs.
This vision builds on subsidiarity: decisions about AI systems should be made at appropriate levels, with local communities controlling systems directly affecting their lives while coordinating at broader scales when necessary. An European city might develop language models trained on local cultural production and historical knowledge, preserving linguistic diversity and regional identity while connecting to broader networks. Professional communities might create specialized models tailored to domain expertise, ethical frameworks, and specialized vocabulary, sharing these tools across their networks while maintaining control over sensitive training data and deployment policies. Regional governments might invest in computational infrastructure and research capacity enabling citizens and institutions to participate meaningfully in AI development rather than remaining dependent on systems developed elsewhere according to foreign priorities and values. Such distributed approaches could foster innovation while maintaining cultural specificity and democratic accountability.
Federated learning offers technical pathways toward this vision. Rather than concentrating training data centrally, federated approaches enable models learning from decentralized sources while preserving local privacy and control. Communities contribute to powerful model training without surrendering data sovereignty, and resulting systems reflect training environment diversity rather than imposing homogeneous templates.
Building distributed cognitive sovereignty requires cultivating European cognitive identity grounded in pluralism, transparency, and digital humanism. This would reflect Europe’s cultural diversity tradition, human rights commitment, and distinctive social market economy. It would resist both purely market-driven American approaches and state-controlled Chinese models, instead embedding AI within broader frameworks of social purpose and human flourishing. Such identity must emerge through inclusive deliberation involving diverse stakeholders in conversations about what intelligence we want to create and what values should guide development.
The future of shared thought
The question facing us is not whether AI will advance, but what governance forms, ownership, and control will shape these systems and to whose benefit they will operate. Will we accept cognitive infrastructure concentrated in few corporations and nation-states, mediated by opaque processes and governed by priorities we do not share? Or can we collectively construct arrangements distributing cognitive power more equitably within democratic accountability frameworks?
Cognitive sovereignty challenges us to recognize that AI governance stakes extend beyond privacy, security, or efficiency. At issue is the future character of human agency itself: our capacity to think freely, reason independently, and make autonomous choices in a world mediated by intelligent systems we do not control. When our knowledge access depends on algorithmic curation we cannot inspect, communications shaped by language models trained on unspecified data, and decisions influenced by opaque recommendations, boundaries between human and machine cognition blur in ways demanding careful governance.
In coming decades, freedom will be measured increasingly by the right to think with our own machines, exercising meaningful sovereignty over cognitive tools mediating our reality relationship. This requires not only technology access but capability to understand, modify, and govern systems according to collectively defined values. It demands new literacy forms enabling citizens to critically evaluate intelligent systems structuring their lives, and new institutional arrangements creating spaces for collective deliberation including diverse voices rather than remaining technical elite provinces.
As we navigate AI’s complex landscape, consider what cognitive sovereignty means in our context. What agency do we wish to exercise over intelligent systems mediating our work, learning, and social interactions? What values should guide technologies increasingly shaping how we think, what we know, and who we become? What communities and practices might you help build expanding distributed cognitive sovereignty possibilities, creating spaces where human and machine intelligence flourish together serving broader purposes rather than narrow interests?
In a world where minds are distributed between humans and algorithms, where cognition itself becomes a shared process transcending boundaries between natural and artificial intelligence, your cognitive sovereignty form will depend on choices we make collectively about governing, developing, and deploying these powerful technologies. The future of thought itself hangs in the balance, and the time to engage is now, before power patterns in this new cognitive landscape become too entrenched to transform.
What is your opinion on this topic?
(Service Announcement)
This newsletter (which now has over 6,000 subscribers and many more readers, as it’s also published online) is free and entirely independent.
It has never accepted sponsors or advertisements, and is made in my spare time.
If you like it, you can contribute by forwarding it to anyone who might be interested, or promoting it on social media.
Many readers, whom I sincerely thank, have become supporters by making a donation.
Thank you so much for your support!
It's interesting how you connect sovereinty with cognition and AI. What if this cognitive sovereignty eventually extends beyond human control, to the AI itself?