Artificial intelligence regulation and Trump’s war with Anthropic

The demand for the regulation of Artificial Intelligence appeared straightforward: systems of increasing capability and complexity ought not to be left to private ambition or technical momentum alone.

Artificial Intelligence needs regulation

Limits, procedures and accountability must be put in place in order to safeguard humanity from the potential risks that unmitigated technological development poses to society. But this procedural language concealed a deeper intuition. There was a growing sense that technological development had slipped beyond ordinary political judgment. The demand for regulation was an attempt to reinsert human deliberation into a process that seemed to be accelerating without reference to any shared horizon.

The idea regulation promised was to give shape to something already unfolding. It was meant to prevent a future in which a small number of firms, backed by venture capital and guided by internal incentives, would determine the conditions of collective life. The state appeared as the necessary counterweight.

Recent events have unsettled this confidence. The confrontation which Novara Media recently described as “Trump’s showdown with Anthropic” has brought into view a different possibility that earlier debates on the ethics of AI only dimly registered.

Anthropic is a US artificial intelligence company founded in 2021 by former OpenAI researchers, organised around the claim that advanced, and possibly independent, systems of advanced intelligence must be designed with caution and direction, with staggering releases to monitor the long-term impacts.

Its work on constitutional training and internal constraint became an image of what responsible innovation, when geared toward social ends, might look like in practice. For many, this seemed to exemplify a future defined by the convergence of technological excellence and public responsibility.

Hegseth v Claude

In January 2026, US defence secretary Pete Hegseth issued a memorandum directing all Department of Defence AI contracts to adopt standardised language permitting “any lawful use” of contracted technologies.

This directive collided with Anthropic’s longstanding usage policies, embedded in its constitutional artificial intelligence framework and acceptable use terms. The company, which had positioned itself as a safety-conscious alternative in the frontier of the AI space, had carved out explicit prohibitions against deploying its models (notably Claude) toward the mass domestic surveillance of US citizens or for application in the development of lethal autonomous weapon systems that require minimal human involvement.

Reports indicated that Claude had already been integrated into classified Pentagon operations, including reportedly aiding in the planning of high-profile actions globally, such as the illegal intervention in Venezuela, which led to the capture of former president Nicholas Maduro.

Nevertheless, when the US government made clear its ambition to integrate artificial intelligence into every layer of military and strategic planning, the limits that Anthropic sought to embed within its systems appeared as an obstruction to the interests of the state.

What had once been celebrated as responsibility began to appear as hesitation in the face of urgency, defined as such from the perspective of the state.

Artificial intelligence as infrastructure

Artificial intelligence is no longer encountered chiefly as a product to be purchased or a tool to be regulated, but as a form of infrastructure upon which the ordering of security, economic life, and knowledge increasingly rests.

Under these conditions, regulation cannot be understood merely as a mechanism for limiting harm. It becomes one of the means through which this emerging order is stabilised, directing the form and distribution of power that can endure within it.

The importance of the present moment lies in this shift of orientation. The institutional and legal arrangements now taking shape will structure how technological progress, political responsibility and authority are understood for years to come.

Their effects will not turn on the outcome of any single dispute, but on the underlying assumptions about risk, security and public accountability that become embedded within them.

Governance no longer stands apart from technological development. It increasingly belongs to the same movement, collapsing the boundaries between regulation as a neutral instrument or tool for state interests.

However, this is not merely a matter of policy design. Instead, it concerns the way artificial intelligence manifests in political life as less a discrete object and more a pervasive condition. Decisions about its deployment shape not only outcomes but the horizon within which future human decisions are made. The field closes and opens at the same time.

Restraint within innovation

In this respect, technology might be seen not only as a collection of tools but a way in which the world discloses itself. What is decisive is not any particular machine, but the manner in which beings become available and ordered. Human capacities are themselves reorganised in response to the way in which artificial intelligence has changed the horizon of human experience.

The conflict with Anthropic makes this transformation visible because it concerns limits. The company’s internal restrictions were attempts to preserve zones in which technological capability would not immediately translate into application. Such limits presupposed that restraint can be maintained within the process of innovation itself.

It was long demanded that private ambition could not be trusted; existential risk required external guardrails; the state must intervene to align technological momentum with human flourishing.

Regulation was envisioned as a necessary corrective to the potentially catastrophic consequences of artificial intelligence, a political form that would temper raw capability with collective foresight. It follows, then, that the demand for regulation was never a simple demand for control, but an attempt to prevent the concentration of power in the hands of a small number of private actors whose decisions could shape the future of humanity exempt from scrutiny.

It would open the process of technological development to scrutiny, to contestation, and to a plurality of interests, interrupting (at least to some extent) the logic of capital.

Public – private collusion

The possibility of nationalising, or effectively absorbing, frontier AI firms such as Anthropic does not so much oppose the power of Silicon Valley as reveal its proximity to the state. The movement between corporate and governmental authority has become increasingly fluid. Expertise, capital and strategic capacity circulate across this boundary. What appears as intervention from outside is often a reorganisation from within.

In this sense, the conflict does not present a simple opposition between technological elites and democratic authority. It exposes the degree to which the political system is already structured through technological infrastructures developed by a narrow social stratum.

The state has become enmeshed in the webs of Silicon Valley, not as a weaver pulling the strings, but as a fly caught in silk spun by others, its every twitch now dependent on the structure of those threads. That is to say, the so-called tech elite does not merely influence power, it constitutes one of its operating centres.

Featured image via the Canary