Should AI access be treated as a civil right across generations?
AI use is expanding faster than the infrastructure that supports it, and that gap is starting to matter for security, resilience, and access.
A new position paper argues that access to AI should be treated as an intergenerational civil right, rather than a service shaped mainly by market forces. The study examines what happens when rising demand for AI collides with limited energy, network capacity, and compute, then proposes a new delivery model to avoid deepening inequality. Even if models keep improving, access to AI outputs will shrink over time unless the underlying architecture changes.
Demand growth collides with physical limits
The paper models how AI inference could scale as it becomes embedded in everyday applications. It assumes peak mobile usage of 60 AI requests per second per user. If about 30 percent of smartphones support AI features, the authors estimate that mobile inference alone could generate more than 5 trillion requests per minute at peak.
That level of demand stresses two constrained resources. The first is networking. Centralized inference points create heavy incast traffic, higher latency, and congestion. Upgrading access and core networks to handle that load is costly and slow.
The second constraint is energy. Prior estimates cited in the paper suggest that a single AI search request can consume up to 1,000 times more energy than a traditional search query. Inference latency targets measured in milliseconds per token push providers toward centralized GPU clusters, which concentrate energy use in a small number of locations.
The authors argue that these pressures make access controls unavoidable. Pricing tiers, usage caps, geographic throttling, and institutional prioritization already exist in limited forms. Over time, these mechanisms become tools to manage scarcity.
Access limits create security and equity risks
The study links access restrictions to security and governance concerns. When AI systems shape education, hiring, healthcare, and research, uneven access translates directly into uneven capability.
The authors warn that selective access undermines merit-based systems. Users with paid or privileged access gain analytical and creative advantages that are unrelated to skill or expertise. This shifts power toward those who can afford faster or deeper AI assistance.
Language coverage and infrastructure gaps amplify the problem. AI systems still favor high resource languages, especially English. Regions with weaker connectivity or energy supply are positioned to receive slower or degraded service even before formal limits appear.
From a regulatory standpoint, the paper notes that existing frameworks do not address this coupling between access and resources. The EU AI Act focuses on risk categories and provider obligations, but it does not establish any right to AI access. US governance relies on voluntary standards and sector rules. UNESCO promotes equitable access, but without binding mechanisms.
Reframing AI as shared infrastructure
The core proposal is to recognize access to AI as an intergenerational civil right. The intent is to protect present day access while preventing resource use that harms future generations.
This framing treats AI as shared social infrastructure, comparable to public libraries or communication networks. AI systems are trained on publicly produced knowledge, including research, text, and cultural material. The authors argue that restricting access to outputs privatizes the benefits of that shared input.
A decentralized AI delivery network
To operationalize this idea, the authors propose an AI Delivery Network, or AIDN. The model borrows from content delivery networks but adapts them for inference rather than static content.
The basic unit is a knowledge fragment, represented as key value cache entries from transformer inference. These fragments can be cached, combined, and reused across the network.
Each AIDN node includes three functions. A storage manager holds local knowledge and metadata. A distribution manager pushes and pulls knowledge fragments based on predicted demand and local context. A generic inference endpoint handles user requests and performs inference when possible, forwarding requests upstream only when needed.
Inference is decomposed across the edge, regional micro data centers, and cloud infrastructure. Lightweight tasks run close to users. More intensive reasoning is invoked selectively. Energy availability, latency, and congestion guide placement decisions.
The authors estimate that inference caching and reuse could reduce compute demand by an order of magnitude for common tasks. This also limits long distance data movement, which the paper identifies as a major driver of energy use.
Access limits and their effect on security operations
Centralized AI services create single points of congestion and policy control. Decentralized inference shifts some risk outward but improves fault tolerance and local control.
The authors position AIDNs as a way to align fairness, sustainability, and operational stability. Without such changes, they argue that AI access will continue to narrow through economic and technical pressure rather than explicit policy choices.
Architecture decisions made now will determine whether AI becomes a durable public capability or a constrained resource available to a shrinking group.
Webinar: Why AI and SaaS are now the same attack surface