New AI Model Shows How a Data Center Moratorium Puts Every American Family at Risk

Anthropic’s newest artificial intelligence (AI) model “Mythos” is already making waves. Anthropic claims the model has found thousands of severe vulnerabilities in major software and operating systems, prompting a partnership with major companies like Microsoft, Amazon, Apple, the Linux Foundation, and more to work on a major cybersecurity initiative dubbed Glasswing.

In the digital age, there are potential gaps in all security infrastructure. This underscores the critical importance of new data center development, which will be needed for the United States to establish and maintain competitive domestic AI capabilities to find and remedy vulnerabilities as its adversaries attempt to exploit them.

The U.S. economy runs on software. In 2024, $368.5 billion was spent on software in the U.S., and those investments fuel everything from agriculture supply chains to healthcare insurance claims and every ATM transaction between. Security flaws like those allegedly identified by Mythos, including zero-day vulnerabilities unrealized for over decades, pose a very real risk to Americans. Glasswing will likely prove an important step toward ensuring that the U.S. is not catastrophically disrupted by mass cybersecurity attacks from malicious actors or governments abroad.

Worries about labor and social disruptions caused by AI have led some to fiercely criticize the technology and propose rules that would cripple its commercial viability. Some state legislatures along with Sen. Bernie Sanders (I-Vermont) have called for moratoriums on data center development. However, even setting aside the overblown nature of these fears, a pause on domestic AI and infrastructure projects would do nothing to stop AI development abroad. Currently, China and the U.S. are locked in a close race for global AI dominance, and a data center moratorium would hand America’s adversaries a major victory in this race and leave the U.S. a sitting duck, vulnerable to cybersecurity risks.

A recent study reported that China was the most prolific country of origin for digital attacks, with Russia and Iran following behind. The actors most likely to engage in cyberattacks against the U.S. are the ones with access to the best AI capabilities outside the U.S. Without heavy investment in the AI ecosystem, the U.S. would be likely to fall behind in the never-ending cybersecurity race and lose the means to play defense against the exponential increase in cyberattacks likely to be deployed against the U.S. from these actors in the near future.

Mythos has ostensibly found previously undiscovered vulnerabilities of existing software. As malicious actors, foreign or domestic, ramp up their capabilities, it will put the existing technology stack at risk. The U.S. could cease all AI operations and shut down all data centers, but Chinese AI models would still be able to exploit American security weaknesses, whether by holding a hospital network hostage or invading a user’s smart TV to invade his home network and gain access to personal information.

Those who wish to impose data center moratoriums are neglecting the very real risks such actions would create. Had such a moratorium been adopted a decade ago, Mythos may not have been a U.S. creation, and the digital vulnerabilities currently being identified and patched could very well have been first found by antagonistic actors abroad. Policymakers must craft technological policy based on the world as it is, not as they wish it was.

 Donald Kimball is the Communications Manager and Tech Exchange Editor for Washington Policy Center, as well as a contributor for Young Voices.