Anthropic announced that its new version "Mythos" could identify and exploit zero-day vulnerabilities in all major operating systems and web browsers when instructed by users. The company explained that its latest artificial intelligence model had found security gaps in the core systems that run nearly all digital infrastructure. This means even amateurs with basic coding knowledge could use these vulnerabilities to hack significant portions of America's digital infrastructure.
Instead of releasing the model publicly, Anthropic formed a consortium with major tech companies including Apple, Google, and Microsoft to begin patching the vulnerabilities. This was a strong signal that the problem was real. It is chilling to consider what might have happened if a corporation or government had reached this technology first. Some will argue that AI should be banned before it steals our passwords and jobs. Unfortunately, as this incident demonstrates, that is impossible. The technology is already out in the world, and if the United States does not develop it, other countries will.
What if this technology had come from a Chinese company? Major Chinese companies are effectively under Communist Party control. This is why China recently issued exit bans on two AI startup founders whose company Meta had acquired. If a Chinese AI developer had discovered massive security vulnerabilities, would they have been allowed to warn the world and distribute patches, or would it have been handed over to China's extensive cyberattack operations?
Some argue for a global development halt through bilateral negotiations. It is an attractive but unrealistic solution. This would be equivalent to major arms control negotiations that take years to decades, yet AI is effectively developing new capabilities every month. Even if diplomatic procedures were accelerated, such treaties remain valid only as long as both sides agree. The United States currently leads in the AI race with a significant advantage over China in "compute," thanks to export controls.
However, China far surpasses the United States in power generation and transmission, another critical component of AI. Over the past four years, China has added capacity equivalent to the entire US power grid. How can the United States trust China to honor any agreement? From China's perspective, it would be more advantageous to agree to a pause while catching up in computing capability, then exit the agreement once the gap closes. This could allow China to cement long-term dominance, as China is expected to close the "computing gap" faster than the United States can fix the legal and regulatory barriers blocking infrastructure development.
All of this points to an uncomfortable conclusion. The United States cannot stop what is coming. Rather than debating a moratorium on new data center construction, it must adapt. Every American company must review its security, and every user must do the same. American politics must channel the passion spent on culture wars into expanding the national power grid. The United States cannot afford endless debates over whether it needs more natural gas or more renewable energy. The answer is "both." An "all of the above" strategy is needed that treats abundant electricity supply as a matter of individual and national defense, not political sparring.
The US government's technological capabilities need a massive upgrade. We do not yet know what AI governance should look like. Strong preemptive rules enacted now would likely stifle development itself rather than promote responsible progress. But it is clear that the government will eventually play a crucial role and is currently unprepared to do so. Defense Secretary Pete Hegseth's recent clumsy attempt related to Anthropic is evidence of this.
Government hiring procedures, pay scales, procurement regulations, and bureaucratic formalities make it impossible to keep pace with private sector expertise. Closing this gap requires sweeping changes across government operations. Federal preemption of state and local regulations is also necessary. State governments lack the technological capacity of the federal government and are unsuited to managing rapidly evolving technology. More importantly, it is not sustainable to proceed with development while coordinating the potentially conflicting and contradictory rules of 50 states.
Until now, most people have thought of AI policy primarily as a domestic issue, a question of how the United States will permit AI development, or whether it will permit it at all. This narrow view obscures the most important question: Will the nation's future be governed by American values and institutions, or by the Chinese Communist Party?
