US AI Policy Shifts to Pre-Verification for National Security

The U.S. AI Policy Shifts Toward National Security Focus

The artificial intelligence (AI) policy of the United States under the Trump administration has undergone a significant transformation, moving from a deregulatory approach to one that emphasizes pre-verification for national security reasons. Initially, the administration prioritized a non-regulatory stance, emphasizing that AI development should not be hindered by excessive regulation. However, as AI capabilities have advanced to the point where they can cause severe damage—such as through cyberattacks or the creation of malicious code—the government has started implementing preemptive management systems to maintain control.

U.S. Government Begins Pre-Screening AI Models

The U.S. Department of Commerce’s AI Standards Innovation Center (CAISI) has signed an agreement with major tech companies such as Google, Microsoft, and xAI to evaluate the performance and security risks of early versions of their AI models. This initiative allows the government to assess potential risks before these models are made public. In 2024, OpenAI and Anthropic had already entered into similar agreements with CAISI. As a result, leading U.S. AI companies are gradually integrating into the government's pre-evaluation system. According to CAISI, over 40 evaluations have already been completed, including assessments of state-of-the-art models that have not yet been publicly released.

The White House is reportedly considering forming a working group to review AI models in advance. According to a report by The New York Times, the administration plans to establish an organization comprising AI company executives and government officials to introduce formal procedures for reviewing new AI models. While the details of this plan have been shared with executives from Anthropic, Google, and OpenAI, it remains unclear whether the initiative will be implemented and which agency will oversee it.

Mythos Shifts U.S. AI Deregulation Stance

This shift marks a complete reversal in the U.S. government’s approach to AI. Shortly after taking office, President Trump abolished the Biden administration’s “AI Safety Executive Order,” adopting a more aggressive strategy to promote AI development. In July of last year, he stated, “This industry (AI) is like a newborn baby. We must nurture and help it thrive,” adding, “It should not be blocked by stupid and foolish rules.” The goal was to position AI as a strategic industry to compete with China on the global stage.

However, the situation changed with the release of Anthropic’s latest AI model, Mythos. Anthropic unveiled Mythos on the 7th of last month, an AI capable of identifying thousands of security vulnerabilities at once. Due to concerns about its advanced capabilities being exploited to threaten cybersecurity systems, access to Mythos was limited to approximately 50 companies and institutions. OpenAI also launched GPT-5.5 Cyber, which matches Mythos’s performance. The White House feared that if such powerful models become widespread, critical infrastructure, government systems, and defense industries could become new targets for attacks. Earlier, the White House opposed Anthropic’s attempt to expand access to Mythos.

Will AI Leadership in the U.S. Shift to the Government?

As AI continues to grow in influence, U.S. government intervention is increasing, mirroring approaches seen in China. Recent defense contracts between U.S. AI companies and the Department of Defense have become a key factor in expanding government influence. On the 1st, the U.S. Department of Defense signed agreements with eight major AI companies, including Nvidia, Microsoft, and OpenAI, to integrate their AI technologies into military networks. This effectively places the Department of Defense in control of how AI technology is used.

The U.S. government is also pressuring companies like Anthropic, which have refused to comply with its standards, by labeling them as “supply chain threat companies.” For AI firms, government contracts offer substantial revenue and technology validation opportunities but also require compliance with government-set security standards and usage conditions.

An AI industry official noted, “As AI, which was intensively nurtured to secure leadership, grew faster than expected, the government is tightening reins,” adding, “AI industry leadership may shift from the private sector to the government.”

Additional Information

For those interested in accessing news and magazines, there are various membership options available. A 5900 won membership offers readers a 2900 won price, while the Chosun Membership provides access to 8 different newspapers and magazines worth up to 55000 won. Members also receive 7000 points that can be used like cash for shopping.