Geostrategic Media Political Commentary, Analysis, Security, Defense
Pranjal Saraswat
Imagine a policy analyst in Brussels arriving at work on an ordinary morning, carrying coffee in one hand and reports on European cyber resilience in the other. Before her meeting begins, she opens an advanced artificial intelligence platform she regularly uses to map cyber vulnerabilities and policy risks, after that she find something has changed i.e. access denied. A revised policy notice flashes briefly across the screen. At first, she feels like it’s some sort of error or technical disruption, but what appears to be a minor inconvenience increasingly resembles a geopolitical warning she wouldn’t think off and that raises a question that no longer feels abstract i.e. what happens when access to artificial intelligence itself begins to require state permission?
That question suddenly feels far less hypothetical after Washington ordered Anthropic to suspend foreign access to its most advanced artificial intelligence models, Mythos 5 and Fable 5, citing national security concerns. US officials feared that the models could be manipulated to identify software vulnerabilities and potentially strengthen cyberattacks, although Anthropic later questioned how serious the risk actually was. Yet if we focus only on the technical dispute, we can miss the broader political significance of the episode. More important than the vulnerability itself was the political reasoning behind the decision. Even liberal capitalist states, which usually support open markets and innovation, often step in when technologies become closely linked to national security. In such cases, governments do not simply leave decisions to private companies but act to protect public safety and strategic interests.
In this view, the state is not merely a regulator but a political actor responsible for protecting citizens and democratic stability when emerging technologies reshape security risks. History offers clear precedents: technologies such as ARPANET and GPS began as strategic military assets before becoming civilian tools, often after periods of restricted access. The pattern is consistent i.e. technologies with strategic value are tightly governed in their early stages. Although artificial intelligence emerged largely through civilian and commercial development, its growing strategic importance is increasingly placing it in a similar category that why it is viewed through the lens of national security and geopolitical competition.
Artificial intelligence, particularly what policymakers and technocrats increasingly call as frontier AI, appears to be entering that phase. Unlike ordinary consumer applications, frontier AI refers to the most advanced generation of systems capable of complex reasoning, cyber modelling, scientific discovery, and strategic analysis. These are not merely tools for productivity or convenience. Their capabilities increasingly touch cybersecurity, intelligence gathering, military planning, and economic competitiveness. This explains why governments increasingly see them less as software products and more as strategic infrastructure.
Publicly, Washington’s move against Anthropic is framed in terms of national security it’s a position that can not be possible without institutional support. Research organizations including RAND and Stanford have warned that highly capable AI systems may improve both cyber offense and defense, accelerate intelligence analysis, and reshape military decision-making. NATO has similarly emphasized artificial intelligence as central to future security planning. Such concerns carry genuine weight. But they do not fully account for what is actually unfolding.
The Anthropic decision points to be something bigger i.e. an emerging arms race which is already reshaping relations between governments and technology firms. Unlike the nuclear competition of the twentieth century, this contest is quieter, largely concealed behind cloud infrastructure, semiconductor supply chains, engineering talent, and export controls. The stakes are not simply about capability. They concern who gains decisive strategic advantage and who shapes the systems that, in turn, determine how security is maintained, economies are run, and critical decisions are made.
This is precisely where the politics become difficult to ignore. In recent years, Washington has increasingly restricted advanced semiconductor exports, especially chips necessary for training sophisticated AI systems. The logic is straightforward: control the computational foundation, and you shape who can compete. The Anthropic restriction extends that logic beyond hardware and into cognition itself. If chips were the first stage of AI geopolitics, advanced models may be the second and that’s the reason US is trying to make law in this regard.
For Europe, this development creates an uncomfortable dilemma. Brussels has invested heavily in regulation, particularly through the European Union’s AI Act, while simultaneously promoting digital sovereignty. Yet Europe remains deeply dependent on American computational infrastructure, cloud providers, and frontier AI firms. This dependence becomes especially significant in light of the Anthropic episode, which suggests that access to advanced AI capabilities may increasingly reflect geopolitical priorities rather than purely commercial relationships. Europe may regulate artificial intelligence, but if frontier systems remain concentrated in American hands, its ability to shape outcomes could ultimately depend on decisions made outside the continent.
The deeper issue that also effectively show as a concern what might be called as computational hegemony or Algorithmic Hegemony which reflect as the concentration of power in the hands of states and big tech that control the infrastructure necessary to develop and deploy advanced artificial intelligence. This includes access to high-performance semiconductors, cloud computing capacity, massive datasets, engineering talent, and the financial resources required to train frontier models. Yet what makes artificial intelligence different from earlier strategic technologies is that resources alone are not enough. Even states or firms with substantial capital and computing power still require the technical skill, organizational capacity, and deployment expertise to transform raw computational resources into functioning frontier systems. In practice, computational hegemony therefore reflects not only who possesses infrastructure, but who can successfully operationalize it at scale — a capability still concentrated among a small number of actors, overwhelmingly based in the United States, who possess disproportionate influence over who can build, access, and benefit from advanced AI systems.
In that sense , geopolitical power has a long-tracked control over strategic infrastructure from maritime routes to energy supplies. So, artificial intelligence may be the latest and most consequential instance of that pattern. According to Stanford’s AI Index, training advanced models now costs tens or even hundreds of millions of dollars, concentrating that power among only a handful of governments and firms capable of sustaining such investment.
The Anthropic episode therefore matters for reasons extending far beyond one company or one disputed vulnerability. It may represent the clearest sign, yet that frontier AI is moving from open technological competition toward geopolitical permission. The analyst in Brussels who found her access revoked one morning was not encountering a technical glitch , she was encountering the early architecture of a new kind of gatekeeping, one where the defining question is no longer what a system can do, but who is permitted to use it, and on whose terms.
About the Author:
Pranjal Saraswat is a PhD Scholar in Gandhian Thought and Peace Studies at Central University of Gujarat-India. Researching on cyber issues, artificial intelligence, and emerging technologies in global politics. His work focuses on the intersection of geopolitics, cyber security, and technological governance. He can be reached out at LinkedIn – www.linkedin.com/in/pranjal-saraswat-2a0b43186
You must be logged in to post a comment.