Technology as an Instrument of the Exercise of Power
The decision of the United States Government rests on the premise that Fable and Mythos enabled the circumvention of safeguards, allowing access to information concerning vulnerabilities in software. Anthropic firmly contests the significance of this risk, noting that comparable capabilities are offered by other publicly available artificial intelligence systems. The company itself emphasizes that it adopted a defense in depth strategy: rather than pursuing an unattainable standard of absolute resilience, it sought to make every circumvention of safeguards as costly and as easily detectable as possible.
Independent of the technical assessment of this dispute, what matters here is the underlying legal mechanism: a single state may, through an administrative decision, deprive users across the entire world of access to a given technology. This is not merely a problem of digital security. It is a question of who exercises control over the digital infrastructure of the twenty-first century, and on what legal basis that control is exercised.
International Law Catching Up
International law has, for decades, attempted to keep pace with digital transformation. As in the case of the natural environment or outer space, this is a domain that escapes the authority of any single actor while simultaneously being subject to regulation by all. Concepts such as jurisdiction, state responsibility, and due diligence acquire new meanings in cyberspace.
The Fable and Mythos case reveals, in this respect, a significant gap: the absence of transparent, binding procedures according to which a state may restrict access to a technology of global reach. Anthropic states explicitly that the government's decision did not meet requirements of transparency and fairness, and that it lacked grounding in technical analysis. This is a serious allegation. In other fields of international law, such as trade law and environmental law, appellate mechanisms and procedural standards have been developed that protect parties against arbitrary exercises of authority. In the field of artificial intelligence, such mechanisms are still lacking.
The appropriate response should be the construction of multilateral regulatory frameworks that combine states' competences in matters of national security with procedural guarantees for entities affected by the decision of a single government. A useful model here may be found in the mechanisms of the World Trade Organization, which permit restrictions on trade for security reasons, but require that such restrictions be justified and subject them to review by independent bodies.
Corporate Responsibility and Human Rights
Technology companies operate globally, yet they are seated in a particular state, are subject to its law, and implement the decisions of its authorities. In light of the United Nations Guiding Principles on Business and Human Rights, a question arises: does a company have an obligation to warn its users in advance? Does an abrupt withdrawal of access violate their legitimate expectations? Does access to digital tools form part of the right to participate in scientific, cultural, or economic life?
The answer is yes, at least in part. The UN Guiding Principles impose on businesses an obligation to exercise human rights due diligence, which also encompasses an obligation to inform users of risks associated with the use of their services. If a company knows that its model may be withdrawn as a result of a government decision, it should disclose this fact in advance. Transparency is not merely good practice; it is an element of the responsible governance of digital infrastructure.
Sustainable Development in the Digital World
The 2030 Agenda envisions technology as a driver of sustainable development, not an obstacle to its realization. If millions of users can be deprived of access to advanced artificial intelligence tools overnight by the decision of a single government, the prospects for achieving goals relating to education, innovation, and the reduction of inequality diminish accordingly. Developing states and those with lesser bargaining power, such as Poland, are confronted with a fait accompli, without any possibility of participating in the decision-making process.
The solution lies in building regional capacity and investing in technological diversification. Europe should not render its research and educational infrastructure dependent on individual providers located outside its borders. Initiatives such as European cloud computing infrastructure, open language models, and common regulatory frameworks for artificial intelligence respond to precisely this risk. The aim is not isolation, but the joint construction of resilience.
What Does This Mean for Internet Users?
The Anthropic case is not an isolated incident. It is a manifestation of a complex process in which law, technology, and geopolitics are becoming ever more closely intertwined. Answers to the questions it raises will not emerge on their own. They require the cooperation of lawyers, technologists, policymakers, and civil society.
The Lodz Cyber Hub, a center for research on international law in cyberspace, follows these developments closely, since they do not unfold solely in Washington or Silicon Valley; the decisions taken there bear directly upon companies, universities, and citizens in Poland and across Europe. The question is not only what artificial intelligence is permitted to do, but above all who decides this, according to what rules, and what accountability they bear toward all those whom the decision affects.
