Beyond the Code: Power, Politics, and the Future of AI

by

Author: Denis Avetisyan

A new analysis argues that focusing solely on AI’s technical risks obscures the crucial economic and political forces driving its development and deployment.

This review deconstructs the ‘political economy of AI’ to reveal how network power and regulatory capture shape the pursuit of accountability and safety.

Despite growing scrutiny of artificial intelligence, efforts to ensure fairness and accountability are often sidetracked by superficial critiques that inadvertently reinforce existing power structures. This paper, ‘Reckoning with the Political Economy of AI: Avoiding Decoys in Pursuit of Accountability’, examines how the development of AI operates as a world-building project, sustained by networks of wealth and power that benefit from strategically deployed distractions. We argue that meaningful progress requires moving beyond technical fixes to confront the underlying political economy of AI and the ways in which its emergent properties are actively constructed. By recognizing these ‘decoys’, can we begin to envision – and build – a more just and technologically entangled future?

The Architecture of Control: AI as World-Building

The pursuit of artificial intelligence extends far beyond the creation of sophisticated algorithms and intelligent machines; it represents a large-scale project of world-building, akin to designing the rules and infrastructure for a future society. This endeavor isn’t neutral; it inherently involves the exercise of power, as those who define the goals, parameters, and deployment of AI systems effectively shape the world according to their vision. The technologies themselves aren’t simply tools, but active agents in constructing new social realities, potentially reinforcing existing inequalities or creating entirely new forms of control. Consequently, analyzing AI necessitates examining not just its technical capabilities, but also the underlying assumptions, vested interests, and power dynamics that guide its development and implementation, revealing how this ‘project’ actively forges a future landscape.

The development of artificial intelligence isn’t a neutral pursuit; it’s deeply embedded within a specific political and economic framework that prioritizes financial gain and centralized control. Current AI infrastructure relies on massive data collection and computational power, resources largely concentrated in the hands of a few powerful corporations and governments. This concentration isn’t accidental, but a consequence of an economic system that incentivizes privatization and profit maximization, effectively creating barriers to entry and limiting broad access to the benefits of AI. Consequently, the narrative of AI as a universally beneficial technology often obscures the reality of its development – a process driven by the consolidation of power and wealth, rather than equitable distribution or societal good. This fundamental reliance on a particular economic logic shapes not only how AI is built, but for whom, and with what ultimate purpose.

A critical examination of artificial intelligence development reveals a discrepancy between its publicly stated ambitions and the forces truly driving its trajectory. The narrative surrounding AI often emphasizes innovation, progress, and solutions to global challenges; however, a deeper analysis of the systemic construction underpinning this field exposes a prioritization of financial gain and centralized control. This isn’t to suggest that beneficial applications are nonexistent, but rather that the pursuit of these advantages is frequently secondary to the accumulation of power and capital. Consequently, understanding the political and economic structures shaping AI isn’t simply an academic exercise; it’s essential for discerning the true motivations behind its development and anticipating its potential societal impacts, as the proclaimed goals often serve as a veil for more fundamental, and potentially problematic, agendas.

Assembling the Infrastructure: The Materiality of Power

The Material Political Economy of Artificial Intelligence is concretely embodied in its Networked Infrastructure, which encompasses the physical hardware – including semiconductors, servers, and data centers – alongside the software systems, algorithms, and crucial data pipelines that enable AI model training and deployment. This infrastructure isn’t a singular entity but a complex, interconnected system requiring substantial capital investment and ongoing maintenance. The production of this infrastructure relies on geographically dispersed supply chains, specialized manufacturing processes, and significant energy consumption. Furthermore, access to and control over these foundational components – from chip fabrication facilities to large-scale datasets – represent key determinants in the current landscape of AI development and influence the distribution of its benefits and risks.

The AI infrastructure is not a neutral collection of technologies, but rather a deliberately constructed ‘assemblage’ comprising hardware components, software algorithms, datasets, and the labor that maintains them. This assemblage is strategically designed and implemented, meaning choices regarding its composition – which technologies are prioritized, what data is used for training, and how resources are allocated – are not made randomly. These choices systematically favor existing power structures by concentrating control and benefits in the hands of specific actors, potentially exacerbating inequalities and limiting broader access to the benefits of AI development. The resulting system reflects and reinforces pre-existing social, economic, and political hierarchies embedded within its design and deployment.

The development of AI is fundamentally reliant on the global circulation of capital, data, and specialized expertise. Financial investment concentrates in regions with established technological infrastructure, primarily North America and East Asia, driving the expansion of computing power and AI research. Simultaneously, large datasets used for training AI models are often sourced from diverse global locations, frequently involving data extraction and processing that occurs outside of the regions where AI systems are ultimately deployed. Crucially, a limited pool of highly skilled AI researchers and engineers, often migrating between international hubs, concentrates knowledge and control over AI development, creating dependencies and imbalances in the global AI landscape. These flows are not simply logistical; they actively shape the priorities, capabilities, and ultimately, the control of AI technologies.

The Art of Obfuscation: Decoys and the Control of Perception

The implementation of ‘Decoys’ represents a deliberate strategy employed by those directing the ‘Project of AI’ to manage public perception and reinforce their position of authority. These tactics involve the calculated dissemination of misleading information and the strategic framing of narratives to distract from the underlying objectives and power consolidation efforts. Rather than addressing substantive concerns regarding AI development, these decoys function to redirect focus, preempt critical analysis, and ultimately, legitimize actions that might otherwise be subject to scrutiny. The consistent application of these diversions allows for the circumvention of accountability and the maintenance of control over the trajectory of AI implementation.

The deployment of decoys within the ‘Project of AI’ manifests as specific rhetorical strategies. The ‘Inevitability Decoy’ presents continued AI development as an unavoidable process, discouraging critical evaluation of its direction or potential consequences. Simultaneously, the ‘Disruption Decoy’ emphasizes the transformative change promised by AI, thereby diverting attention from the consolidation of power occurring alongside its implementation. This tactic frames the narrative around innovation while obscuring the underlying mechanisms of control and accountability, effectively precluding scrutiny of who benefits from, and governs, these advancements.

The ‘Ontological Decoy’ and ‘Regulatory Decoy’ operate in tandem to shape discourse surrounding AI development and preemptively justify existing structures of power. The Ontological Decoy achieves this by framing AI as an inherent, almost inevitable, force – a new ontological reality – thereby shifting the focus from how AI is developed to simply accepting that it is being developed. Simultaneously, the Regulatory Decoy focuses on creating the appearance of oversight through often superficial or easily circumvented regulations. This combined approach preemptively addresses potential criticisms by defining the boundaries of acceptable debate and legitimizing the actions of those already in control, effectively preventing meaningful accountability for the development and deployment of AI technologies.

Networked Control: The Illusion of Progress and the Erosion of Systemic Change

The foundational element driving the advancement of artificial intelligence isn’t simply technological innovation, but rather a dynamic known as ‘Network Power’. This refers to the capacity to shape outcomes not through direct command, but through the intricate web of relationships between researchers, corporations, policymakers, and even public perception. Influence isn’t held by a single entity, but circulates throughout these connections, allowing priorities and development paths to be subtly guided. Those who strategically cultivate and leverage these interconnected relationships – funding research, shaping narratives, and securing regulatory advantages – effectively wield disproportionate control over the future of AI, often prioritizing specific applications and overlooking potential societal consequences. The strength of this network power lies in its decentralization; it’s a system where influence is exerted through collaboration and mutual benefit, making it both robust and difficult to challenge.

The pursuit of artificial intelligence often centers on mitigating immediate technical risks, a phenomenon described as the ‘Safety Decoy’. While crucial, this intense focus on issues like algorithmic bias and unintended consequences can inadvertently overshadow the larger societal implications of increasingly powerful AI systems. By prioritizing narrowly defined safety parameters, attention and resources are diverted from examining the potential for job displacement, the exacerbation of existing inequalities, and the concentration of power within the hands of those controlling these technologies. This isn’t to suggest technical safety is unimportant, but rather that an overemphasis on it creates a convenient deflection, allowing developers and policymakers to avoid confronting the more challenging, systemic questions surrounding the true impact of AI on society and the distribution of its benefits.

The pervasive narrative of ‘Solutionism’ – the belief that technological advancements inherently provide solutions to all societal challenges – actively diminishes the crucial need for broader systemic change. This perspective, frequently amplified within the discourse surrounding artificial intelligence, suggests that complex problems like inequality, climate change, or political polarization can be resolved through innovative apps, algorithms, or automated systems. While technology can undoubtedly play a role in addressing these issues, an overemphasis on technological fixes obscures the underlying social, economic, and political factors that contribute to them. Consequently, resources and attention are often diverted from addressing root causes, fostering a cycle where technological ‘solutions’ treat symptoms rather than tackling fundamental problems, and reinforcing the status quo under the guise of progress. This creates an illusion of advancement, masking the necessity for comprehensive, multifaceted change that extends beyond the realm of technology.

The study of artificial intelligence, much like any complex system, reveals a tendency toward assemblage-a bringing together of disparate parts driven by underlying forces. It is not simply about building intelligent machines, but about the construction of a future shaped by economic and political power. As Edsger W. Dijkstra observed, “It’s not about understanding everything, it’s about understanding what’s important.” This sentiment echoes the paper’s central argument: that focusing on the ‘political economy of AI’-the power dynamics and economic forces at play-is crucial. To attempt to address AI safety without acknowledging these underlying structures is to chase decoys, diverting attention from the true mechanisms of control and the construction of a specific, potentially inequitable, future.

What Lies Ahead?

The analysis presented here does not offer solutions, because the problems it identifies are not malfunctions to be fixed, but inherent conditions. The pursuit of ‘AI safety’ framed as purely technical risk mitigation risks becoming a palliative measure, obscuring the deeper entrenchment of existing power structures. Every bug, in this light, is a moment of truth in the timeline – a fleeting glimpse of the interests coded into the assemblage. The question is not whether these systems can fail, but how they will inevitably reflect the biases and incentives of their creators and deployers.

Future inquiry must move beyond deconstructing the technical facade and actively map the flows of capital, influence, and regulatory capture that constitute the political economy of AI. The focus should shift from anticipating catastrophic failure to understanding the more subtle, continuous processes by which these systems shape – and constrain – possible futures. This necessitates a willingness to confront the uncomfortable truth that technical debt is the past’s mortgage, paid by the present, and accruing interest for generations to come.

Ultimately, the longevity of any approach to AI governance will not be measured by its immediate effectiveness, but by its capacity to age gracefully – to adapt and reveal, rather than conceal, the inevitable decay inherent in all complex systems. The real challenge lies not in preventing change, but in understanding and navigating its consequences.

Original article: https://arxiv.org/pdf/2604.16106.pdf

Contact the author: https://www.linkedin.com/in/avetisyan/

See also:

2026-04-20 08:37