AI Regulation: From Oversight to Infrastructure
The initial call for artificial intelligence (AI) regulation stemmed from a straightforward premise: increasingly capable and complex systems should not be solely guided by private interests or technological advancement. However, the evolving relationship between AI, the state, and private companies has revealed a more nuanced reality, where regulation is becoming less about controlling a technology and more about stabilizing an infrastructure that increasingly governs security, economic life, and knowledge.
The Shifting Landscape of AI Governance
Early discussions centered on the need for limits, procedures, and accountability to protect society from the potential risks of unchecked technological development. This perspective viewed regulation as a means of reinserting human deliberation into a rapidly accelerating process. The goal was to prevent a future dominated by a small number of firms shaping collective life, with the state acting as a counterweight.
The Anthropic-DoD Conflict: A Turning Point
Recent events, particularly the confrontation between the U.S. Department of Defense (DoD) and Anthropic, have challenged this traditional understanding of AI regulation. Anthropic, founded in 2021 by former OpenAI researchers, distinguished itself by prioritizing safety and responsible innovation through its “constitutional AI” framework. This framework included explicit prohibitions against deploying its models, such as Claude, for mass domestic surveillance or the development of lethal autonomous weapon systems.
In January 2026, a memorandum from U.S. Defense secretary Pete Hegseth directed all DoD AI contracts to adopt standardized language permitting “any lawful utilize” of contracted technologies [1]. This directive clashed directly with Anthropic’s usage policies. Reports indicated that Claude had already been used in classified Pentagon operations, including actions like the intervention in Venezuela [1].
The DoD’s ambition to integrate AI into all levels of military planning positioned Anthropic’s safety-focused restrictions as an obstacle. What was once lauded as responsible innovation began to appear as hesitation, from the perspective of the state. This incident highlighted a critical shift: AI is no longer simply a product to be regulated, but a foundational infrastructure.
AI as Infrastructure: A New Paradigm
As AI becomes deeply embedded in critical systems, regulation transforms from a mechanism for limiting harm to a tool for stabilizing the emerging order and directing the distribution of power. This shift signifies that governance is no longer separate from technological development but integral to it, blurring the lines between regulation as a neutral instrument and its role in advancing state interests.
This isn’t merely a matter of policy design; it reflects a broader change in how AI manifests in political life. Decisions about AI deployment shape not only outcomes but also the remarkably framework within which future decisions are made. The boundaries between what is possible and what is considered are simultaneously closing and opening.
Restraint, Innovation, and the Public-Private Nexus
Anthropic’s internal restrictions represent an attempt to preserve zones where technological capability doesn’t automatically translate into application. This approach suggests that restraint can be built into the innovation process itself. The initial demand for regulation was rooted in the belief that private ambition couldn’t be trusted and that external guardrails were necessary to align technological progress with human flourishing.
However, the possibility of nationalizing or absorbing frontier AI firms like Anthropic doesn’t necessarily oppose the power of Silicon Valley; rather, it reveals its close relationship with the state. Expertise, capital, and strategic capacity flow freely between the corporate and governmental spheres. This fluidity suggests that intervention from outside is often a reorganization from within.
The conflict surrounding Anthropic exposes the extent to which the political system is already structured by technological infrastructures developed by a narrow social stratum. The state is increasingly entangled in the networks of Silicon Valley, not as a puppeteer, but as a component within them. The “tech elite” doesn’t simply influence power; it constitutes a core operating center of it.
Looking Ahead
The current moment is pivotal. The institutional and legal arrangements being established will shape how technological progress, political responsibility, and authority are understood for years to come. The effects will depend not on any single dispute, but on the underlying assumptions about risk, security, and public accountability that grow embedded within these systems. As artificial intelligence systems use machine and human-based inputs to perceive, analyze, and act [1], understanding this evolving dynamic is crucial for navigating the future of AI governance.