Trump and AI: Between Deregulation and Security Concerns

An Initial Stance Against Strict Regulation

In February 2025, Vice President JD Vance spoke at the Paris AI Action Summit to outline the Trump administration's position on artificial intelligence regulation. He warned against the danger that overly strict regulations could pose to a rapidly growing industry. According to him, these rules could stifle a promising technology just as it begins to take off. Vance also suggested that some AI companies, by calling for regulation, might be seeking to eliminate future competition.

Vance's remarks reflect a widely shared opinion among Trump administration officials that fears surrounding AI capabilities are often exaggerated. David Sacks, the White House's AI and cryptocurrency lead, spoke of a "fear industrial complex" that would employ a "sophisticated fear-based regulatory capture strategy." Michael Kratsios, head of the Office of Science and Technology Policy, criticized international efforts to regulate AI, claiming they foster a climate of fear.

A Lobbying Campaign Against Regulation

The Trump administration has backed its position with a lobbying campaign aimed at preventing the adoption of AI regulations at the state level. According to Axios, administration officials have pressured Republican lawmakers in states like Nebraska and Tennessee to weaken or abandon bills that would impose safety and transparency requirements on AI companies.

What makes this situation particularly notable is that Trump is considering reviving a Biden-era idea he had previously rejected: government pre-review of new powerful AI models.

Towards Government Review of AI Models

Discussions are underway within the administration to establish an executive order creating an AI task force. This group would bring together tech leaders and government officials to examine potential oversight procedures. Among the proposals is a formal government review process for new AI models, similar to what is being developed in the UK, where several government agencies are tasked with ensuring that models meet certain safety standards.

Under the Biden administration, an executive order had been issued requiring AI companies to conduct safety tests and share the results with the government before launching new models. Trump revoked this order early in his second term, replacing these requirements with a new directive aimed at "removing barriers to American leadership in artificial intelligence."

The Mythos Case and Its Security Implications

The Trump administration's shift is partly motivated by Anthropic's Mythos model. This model, currently in a limited preview for select companies, has demonstrated its ability to develop cybersecurity exploits, raising concerns within the government due to potential national security risks. The White House opposes expanding access to Mythos, fearing that the company may not have enough computing power to serve government clients, although Anthropic disputes this claim.

Paradoxically, the Trump administration has also sought to designate Anthropic as a "supply chain risk" due to its refusal to modify a contract with the Pentagon. This designation is being defended in court, while the administration simultaneously works to expand access to Mythos for government agencies.

A Changing Regulatory Environment

The AI industry faces a regulatory environment reminiscent of the Biden era, where companies must submit their models to the government for review before launching them. Google, Microsoft, and xAI recently announced that they would provide the government with early access to their models, with reviews managed by the Center for AI Standards and Innovation at the U.S. Department of Commerce.

Before Trump's arrival, this agency was known as the U.S. AI Security Institute. Its name was changed last June to reflect a less restrictive approach. However, less than a year later, the administration appears to be reassessing its position in light of growing security concerns.

Internal Tensions and International Stakes

The National Security Agency is now using the Mythos model to identify vulnerabilities in Microsoft products, while anticipating that foreign nations could use similar technologies against U.S. critical infrastructure. This situation puts the Trump administration at odds with its own base, as the public increasingly expresses opposition to data centers and other symbols of AI power.

The administration's rhetoric on AI, centered around the idea of "getting rich and beating China," has failed to resonate beyond venture capital circles. As a result, Trump's effort to impose a moratorium on most state-level regulations seems increasingly unfeasible.

Towards International Cooperation or Increased Control?

The diminishing influence of tech acceleration advocates could lead to increased pressure to expand export controls on powerful chips to China. A less likely but desirable development would be for the U.S. to reconnect with allies like the UK, Japan, and South Korea to develop a joint governance strategy for AI models.

However, critics fear that the Trump administration might use a potential licensing regime to censor models deemed "woke" or to exert pressure on companies. This concern is fueled by the possibility that the White House could impose restrictions similar to those of the Federal Communications Commission, but applied to AI.

A year after Vance's speech in Paris, the Trump administration is beginning to take AI security concerns seriously. Models are becoming increasingly powerful and potentially dangerous, and what Sacks once referred to as the "fear industrial complex" now includes a growing number of federal agencies and Trump administration officials. While they should have paid attention to these concerns from the start, it is reassuring to see that they are finally taking them seriously.