National Security Card Pulled On AI

A political figure with a serious expression standing outdoors near the White House

President Trump has signed a new executive order asking artificial intelligence companies to voluntarily submit powerful new AI models to the federal government for review 30 days before public release — a move framed as a national security necessity that also tests the administration’s own commitment to keeping government out of the way of American innovation.

Story Snapshot

Trump signed an executive order establishing a voluntary 30-day government review window for new frontier AI models before public release.
The order is framed around national security and cybersecurity, with agencies like the National Security Agency already examining AI models for vulnerability discovery.
The framework is voluntary, not mandatory — a scaled-back version reflecting Trump’s earlier concerns that heavy-handed regulation could slow America’s lead over China.
The order also directs the Attorney General to challenge state AI laws that conflict with federal policy, centralizing AI governance at the federal level.

Trump Signs Scaled-Back AI Review Order

President Trump signed an executive order establishing a voluntary framework under which developers of frontier AI models are asked to submit new systems to the federal government for review 30 days before public release. The order focuses on national security and cybersecurity applications, giving agencies access to examine powerful models before they reach the public. The framework is deliberately voluntary — a significant departure from earlier, more prescriptive drafts that Trump himself shelved.

Trump previously postponed a stricter version of this order, explaining his concern directly: he did not want any policy getting in the way of America’s technology lead over China and other rivals. That instinct shaped the final order’s voluntary structure. The White House has framed the current approach as a minimally burdensome national framework — consistent with the administration’s broader posture of removing regulatory barriers while still asserting federal authority over AI security matters. ^[5]

National Security Is the Core Justification

The White House has been explicit that United States leadership in AI will promote national and economic security across multiple domains. ^[5] Agencies including the National Security Agency have already been reviewing AI models — specifically examining systems like Anthropic’s Claude for cybersecurity vulnerability discovery. Reporting indicates that advanced AI models have been responsible for uncovering a significant share of recently discovered software vulnerabilities, giving the government a concrete security rationale for early access.

The review framework allows federal agencies to assess frontier models for risks tied to foreign adversaries, critical infrastructure, and cyber threats before those systems are widely deployed. The White House has also reportedly limited broader government access to certain AI models due to foreign-adversary concerns — a sign that national security calculations are driving real decisions, not just providing political cover for regulatory expansion. ^[7]

Preempting State Overreach on AI

Beyond the review framework, the December 2025 executive order takes direct aim at state-level AI regulations that conflict with federal policy. ^[3] The order directs the Attorney General to establish an AI Litigation Task Force within 30 days, with a mandate to challenge state AI laws deemed inconsistent with the federal framework. ^[1] The White House has identified state laws that would require AI models to alter truthful outputs — or impose burdensome compliance regimes — as specific targets for federal preemption. ^[5]

Donald Trump Signs AI Executive Order That Includes Voluntary Review Period For New Models https://t.co/XIhhEBHT89

— Deadline (@DEADLINE) June 2, 2026

This federal preemption strategy reflects a calculated trade-off: centralize AI governance in Washington to prevent a patchwork of conflicting state rules from fragmenting the market and slowing American AI companies. The administration’s AI action plan, published at ai.gov, lists multiple executive orders aimed at accelerating data center permitting, promoting AI exports, and cutting bureaucratic friction. ^[7] Keeping states from imposing their own divergent mandates — including diversity, equity, and inclusion requirements embedded in AI systems — fits squarely within that deregulatory agenda. ^[14] The conservative case here is straightforward: one federal standard beats fifty state experiments in social engineering applied to technology.

Balancing Security and Innovation

The tension in this order is real and worth acknowledging honestly. Trump’s own instinct — that government review could slow America’s 18-month AI lead over China — is a legitimate concern shared by many in the technology industry. A voluntary framework addresses that concern partially, but voluntary frameworks can become mandatory ones over time. Conservatives who rightly distrust government overreach should watch whether this 30-day window remains genuinely voluntary or quietly hardens into a de facto approval process as agencies build institutional momentum around it.

For now, the order reflects a reasonable attempt to balance competing priorities. The United States cannot afford to let adversaries exploit vulnerabilities in AI systems that were never reviewed by anyone with a security clearance. At the same time, the administration’s credibility on limited government depends on keeping this framework narrow, transparent, and genuinely optional. The order’s success will ultimately be measured not by how many models get submitted, but by whether the review process produces real security benefits without becoming another Washington bureaucracy looking for reasons to expand its own authority. ^[8]