The arrival of an AI model known as Mythos has reportedly prompted emergency-level deliberations within the U.S. government, marking what may be a turning point in how state intelligence agencies assess the intersection of machine learning and national security. Details about the model's specific capabilities remain closely held, but the reaction from Washington — described by Oscar Jonsson in Dagens Nyheter as a crisis meeting of senior leadership — suggests that Mythos represents a qualitative shift in automated offensive potential. For middle powers watching from the outside, the episode raises an uncomfortable question: what happens when the tools that define national defense are developed, owned, and controlled elsewhere?

Jonsson, a scholar of Russian warfare writing in Sweden's paper of record, frames the moment as a wake-up call. Nations like Sweden, he argues, face a binary choice — invest in sovereign AI capabilities or accept deepening dependence on the United States for digital security. The framing is deliberately stark, and it arrives at a time when the broader discourse around AI sovereignty has moved from academic abstraction to active policy debate across Europe and the Indo-Pacific.

Automated Offense and the Speed Problem

The strategic concern at the heart of the Mythos discussion is not artificial intelligence in the abstract but a specific asymmetry: the speed of automated offensive tools versus the structural response time of traditional state institutions. Cyber operations have long moved faster than bureaucratic decision-making, but earlier generations of tools still required significant human direction. The emerging class of AI models — of which Mythos appears to be a leading example — threatens to compress the offensive cycle further, potentially enabling reconnaissance, exploitation, and action at machine speed with minimal human oversight.

This is not an entirely new dynamic. The concept of an "automation gap" in cyber conflict has been discussed in defense circles for years, particularly after incidents that demonstrated how quickly automated tools could identify and exploit vulnerabilities across networked systems. What appears to distinguish the current moment is the breadth of capability attributed to frontier models. When a single system can reportedly prompt crisis-level meetings among senior officials of the world's dominant military power, the implication is that existing defensive architectures may be structurally insufficient — not merely underfunded, but conceptually outpaced.

For states that rely on allied infrastructure and intelligence-sharing arrangements for their cyber defense posture, this gap carries a secondary risk. In a crisis that unfolds at machine speed, the time required to coordinate across alliance structures may itself become a vulnerability.

Sovereignty as a Defense Doctrine

Jonsson's argument in Dagens Nyheter situates the Mythos episode within a broader thesis about sovereignty. The traditional markers of national autonomy — territorial control, independent foreign policy, domestic industrial capacity — are being joined by a new requirement: the ability to develop, deploy, and govern advanced AI systems domestically. A nation that cannot independently assess a threat like Mythos, let alone counter it, occupies a fundamentally different strategic position than one that can.

This line of reasoning echoes debates already underway in the European Union, where the AI Act established a regulatory framework but left questions of strategic capacity largely unaddressed. France has pursued a national AI strategy with explicit sovereignty language. The Nordic countries have invested in joint cyber defense coordination. Yet the gap between regulatory ambition and operational capability remains wide. Building sovereign AI infrastructure requires not only funding but sustained access to talent, compute resources, and data — inputs that are currently concentrated among a small number of private firms, most of them American.

The tension, then, is not simply between national autonomy and alliance dependence. It is between the pace at which frontier AI capabilities are advancing and the pace at which mid-sized democracies can build institutional capacity to match. Jonsson frames Mythos as a catalyst — a moment that should force a reimagining of cyber defense as a core pillar of sovereignty rather than a technical annex to traditional military planning.

Whether that reimagining actually occurs depends on factors that remain unresolved: political will, budget allocation, and the willingness of allied governments to share rather than hoard the most sensitive capabilities. The Mythos episode has surfaced the question with unusual clarity. The answer, for Sweden and states in similar positions, is far less clear — and the window for deliberation may be shorter than any of them expect.

With reporting from Dagens Nyheter.

Source · Dagens Nyheter