Who Controls the Cognitive Infrastructure? The Real Stakes Behind Mythos and Tulongfeng

The strategic competition between the United States and China is often framed as a race to build the most powerful artificial intelligence model.

But developments in recent weeks point to something more complex, and arguably more consequential.

The real contest is not simply about who owns the most advanced model, but about who can best weave together artificial intelligence, data, infrastructure, and operational capability into a working whole.

The unveiling of Tulongfeng, the new system built by Qihoo 360, was cast as China’s answer to Mythos, Anthropic’s automated vulnerability-research platform. Setting the two systems side by side offers a useful window into how global technological competition is evolving.

For years, debate over artificial intelligence centered on the power of language models themselves — their size, their parameter counts, their reasoning ability.

A different paradigm is now taking shape.

According to Qihoo 360, Tulongfeng does not rest on a single dominant language model. Instead, it runs on a collection of specialized agents that work together, each handling a distinct piece of the task: some map attack surfaces, others trace information flows, others build test environments, confirm vulnerabilities, or generate the exploit code needed to act on what they find.

The logic echoes a shift that has already reshaped Western military thinking. For decades, military superiority was measured by the quality of individual platforms — tanks, ships, fighter jets. Today it depends increasingly on the ability to knit sensors, networks, command systems, artificial intelligence, and human operators into a single distributed cognitive architecture.

That same transformation now seems to be arriving in the cyber domain. The center of gravity is no longer the individual model. Advantage comes not from any one system’s quality but from the ability to integrate and coordinate a network of specialized intelligent agents, much as sixth-generation fighter programs no longer rest on the platform alone but on the network linking sensors, systems, and decision-makers.

The United States still holds a clear lead in so-called “frontier models,” and companies like OpenAI, Anthropic, and Google continue to set the global benchmark for advanced AI research.

China, however, does not appear to be chasing frontier-model leadership alone. Its ambition looks broader: to fold artificial intelligence, data, digital infrastructure, and operational experience into a single ecosystem capable of converting technological innovation into power.

That is where the deeper geopolitical question emerges. The competition is not only about who builds the most advanced technology, but about who can fold it fastest into their economic, military, and decision-making structures. Seen this way, systems like Mythos and Tulongfeng are far more than software tools. They mark the arrival of a new kind of strategic capability: the power to identify, analyze, and exploit digital vulnerabilities at a speed and scale no human team could match alone.

The dual-use character of these technologies is plain to see. They can harden critical infrastructure, strengthen the resilience of digital systems, and bolster national security. Just as easily, they can become instruments of competition, surfacing an adversary’s vulnerabilities, laying the groundwork for offensive operations, and yielding an edge in the cyber domain. Defense and offense begin to blur together.

None of this is new in kind; military history is full of dual-use technology. What is new is how much artificial intelligence multiplies the speed, scale, and impact of it. That is why this cannot be dismissed as ordinary industrial competition among tech companies. It is a transformation bearing directly on the redistribution of power among states. Whoever controls the cognitive infrastructure of the future will hold a growing edge in defending networks, gathering intelligence, planning operations, and managing complexity.

In a world only growing more interconnected, the ability to understand, anticipate, and steer complex processes may become one of the primary currencies of power. Artificial intelligence, in other words, is evolving from an enabling technology into an infrastructure of power in its own right.

What the comparison between Mythos and Tulongfeng really reveals goes well beyond a rivalry between two technical systems. It marks the emergence of a new cognitive arms race, one in which the deciding factor will not be technological superiority alone, but the ability to fuse human talent, algorithms, data, and networks into ever more sophisticated cognitive architectures.

As with sixth-generation aircraft programs, the center of gravity is shifting away from the platform itself and toward the cognitive network binding every element of the system together.

It is likely that a significant share of the coming redistribution of international power will be decided in that transition.

This is translated from the Italian and the original article published in European Affairs on 3 July 2026 and is published with the permission of the author.