The Trump administration's approach to artificial intelligence has become so internally contradictory that it now undermines its stated goal of maintaining American dominance in the global AI race. Just hours before a planned executive order signing ceremony in May, the president scrapped the agreement after complaints from White House advisers. Then, on June 3, he signed a different executive order that creates a voluntary process for AI companies to share upcoming models with the government for safety testing up to one month before public release.

What's Actually in Trump's New AI Executive Order?

The executive order requires major AI firms like OpenAI and Anthropic to voluntarily share certain upcoming models with federal agencies for safety review and to collaborate on strengthening cyberdefenses. However, the policy is largely symbolic. "The new rule effectively formalizes what has already been happening between the U.S. government and the leading AI companies," explained Daniel Remler, an AI expert at the Center for a New American Security.

The order itself celebrates Trump's earlier decision to rescind Biden-era AI regulations, calling them "dangerous" and a "barrier" to American AI leadership. Yet the core components of those supposedly harmful Biden policies, such as voluntary agreements to share information about advanced AI models with federal agencies, are strikingly similar to what Trump just signed. "The policy is considerably more intrusive" than Biden's executive order, according to Dean Ball, a former AI adviser to the Trump administration.

Why Is the Administration Sending Mixed Signals on China?

The contradictions extend far beyond domestic regulation. Trump has repeatedly emphasized the need to deregulate the AI industry to stay ahead of China in the global AI race. Yet simultaneously, his administration has permitted Nvidia to sell some of its most advanced AI chips to Chinese companies, lifting export controls that the Biden administration had specifically imposed to slow Chinese AI development.

The administration's stance on Anthropic, one of the leading AI companies, further illustrates the incoherence. In April, Anthropic announced Claude Mythos Preview, a model with advanced hacking capabilities that triggered concern over AI power. Trump appeared to embrace the company afterward, telling reporters, "I like high-IQ people, and they definitely have high IQs" when discussing Anthropic's leadership. Yet the Pentagon designated Anthropic a "supply chain risk" in February, essentially declaring it a national-security threat for the military to use its products. When Anthropic tried to grant Mythos access to more companies for cyberdefense work in late April, the White House appears to have blocked the move, despite this aligning with the administration's stated cybersecurity goals.

In April, Anthropic

How Is the Administration Undermining Its Own Cybersecurity Goals?

Perhaps most troubling, the Trump administration has gutted the Cybersecurity and Infrastructure Security Agency (CISA), the federal agency tasked with protecting the nation against cyberattacks. CISA is also one of the main agencies responsible for implementing today's executive order on AI safety and government-industry collaboration. This creates a fundamental contradiction: the administration is asking AI companies to strengthen federal cyberdefenses while simultaneously weakening the agency that coordinates that defense.

The administration's energy policy reveals similar inconsistency. Trump has pushed to remove regulatory constraints on data-center construction to maintain American AI leadership, declaring "Build, baby, build" last July. But once public concern emerged about data centers driving up electricity bills, the White House announced a voluntary pledge for AI companies to take measures preventing everyday people from paying for data-center electricity.

Steps to Understanding the Policy Landscape

• Voluntary vs. Mandatory: The new executive order relies entirely on voluntary cooperation from AI companies rather than enforceable regulations, making compliance dependent on industry goodwill rather than legal obligation.
• Domestic Regulation vs. Export Policy: The administration simultaneously restricts AI development domestically through safety reviews while loosening export controls that prevent advanced chips from reaching Chinese competitors.
• Cybersecurity Rhetoric vs. Agency Cuts: The order emphasizes strengthening federal cyberdefenses while the administration has reduced funding and authority for the primary agency responsible for implementing those defenses.

The chaotic policy trajectory reflects deeper political pressures. Over recent months, as AI models have improved, public attitudes toward the technology have soured. AI spending is consuming the U.S. economy, workers fear job displacement, and communities across the nation are protesting data centers. Political figures as divergent as Steve Bannon, Trump's former chief strategist, and Bernie Sanders have expressed concern over AI concentration of power.

"It's not perfect. But directionally, it is pretty damn good," said Steve Bannon regarding the executive order.

Steve Bannon, Former Chief Strategist

The administration appears to be attempting to satisfy both the public and Silicon Valley simultaneously. By signing an executive order, Trump can claim he is undertaking robust AI regulation. Yet by keeping the requirements voluntary, he avoids imposing meaningful burdens on the tech industry that has become a major economic and geopolitical power. This middle path, however, may satisfy neither constituency.

The window for any single government to seriously regulate AI is rapidly closing as companies like Anthropic, OpenAI, and their competitors become major economic and geopolitical powers. The Trump administration's contradictory approach, driven by competing political pressures rather than coherent strategy, may have already squandered the opportunity for meaningful American AI governance.