Trump Administration Shapes AI Industry Through Ad-Hoc Interventions Despite Deregulation Rhetoric

**An Anti-Regulation Agenda That Regulates Anyway**

The Trump administration came into office with a clear ideological commitment: step back, clear the runway, and let American AI companies build without the friction of federal oversight. It revoked Biden-era executive orders that had imposed reporting requirements and safety assessments on frontier AI developers, and it positioned itself in stark contrast to the European Union's sweeping AI Act and the United Kingdom's Online Safety Act. The message was deregulatory, emphatic, and politically consistent with broader Republican instincts about government interference in private industry.

The reality on the ground has turned out to be considerably more complicated. Rather than stepping aside, the administration has become deeply enmeshed in the AI industry's trajectory — not through formal rulemaking or legislation, but through a series of case-by-case interventions that amount to an informal, ad hoc governance regime. Without Congress passing any comprehensive national AI legislation, the executive branch has effectively become the primary force shaping how the most powerful technology in the world gets built, deployed, and exported. The practical effect is a kind of shadow regulation — consequential, directional, and often opaque.

This matters not just domestically but globally. Because the United States is home to the most capable frontier AI systems, Washington's decisions carry disproportionate weight. What the administration chooses to permit, restrict, or encourage cascades outward into decisions made by companies, governments, and researchers in dozens of other countries. The absence of explicit rules does not mean an absence of influence — it may, in fact, amplify it, since companies must constantly second-guess what the administration wants rather than following a codified set of requirements.

**The Mechanisms of Informal Control**

The clearest example of this dynamic involves export controls. Anthropic, the San Francisco-based AI safety company backed by billions in investment, is currently in active negotiations with the administration over whether export restrictions can be lifted on its latest models. The outcome of that conversation will determine which foreign governments and companies can access Anthropic's most capable systems — a decision with profound geopolitical and commercial implications. Yet it is being resolved not through a published regulatory framework or a notice-and-comment rulemaking process, but through a bilateral negotiation between a private company and executive branch officials.

Anthropic is not alone in navigating this landscape. Other leading AI laboratories — including those developing systems that rival or may soon surpass current frontier capabilities — are working to understand what compliance looks like under a recent executive order that established a voluntary framework for government review of advanced AI models. The word "voluntary" is doing significant work in that formulation. When the world's most powerful government signals it wants to review your models, the voluntariness of that arrangement is at best nominal. Labs are effectively operating under a review regime without the procedural protections or predictability that formal regulation would provide.

The administration has also directed significant energy toward preempting state-level AI laws. A number of states — including California, Colorado, and Texas — have moved to fill the federal vacuum with their own AI governance measures. The administration's opposition to these efforts, framed as protecting a coherent national approach, has the paradoxical effect of clearing space for federal influence while simultaneously refusing to codify that influence as law. The result is a patchwork-prevention strategy that leaves the administration as the de facto, if unofficial, national AI regulator.

**Constitutional Ambiguity and Democratic Accountability**

There is a structural tension embedded in all of this that goes beyond any single policy decision. Under the American constitutional system, it is Congress's role to write laws, and the executive branch's role to enforce them. When the administration shapes industry behavior through executive orders, procurement decisions, export control negotiations, and informal pressure — rather than through legislation — it is exercising a form of governance that bypasses the normal mechanisms of democratic accountability. There are no public comment periods, no legislative hearings, no formal records of how decisions are reached.

This creates real uncertainty for AI companies trying to plan multi-year investments and product strategies. A framework that exists through executive discretion can change with a phone call, a personnel shift, or a new geopolitical irritant. The voluntary government review process, for instance, provides no clear standards for what triggers review, what criteria are applied, or what remedies are available if a company disagrees with an outcome. For startups and international competitors watching from the outside, the rules of engagement are genuinely unclear.

The administration's approach also raises questions about where accountability lands when things go wrong. If a powerful AI model causes harm — through misuse, accident, or unforeseen capability — the absence of a formal regulatory record makes it difficult to reconstruct what oversight existed and whether it was adequate. Informal governance works well when it works; it provides little recourse when it doesn't.

**Why the Rest of the World Is Watching**

The global stakes of this ambiguity are substantial. The EU's AI Act, now moving into implementation, creates binding obligations for any company offering high-risk AI systems in European markets — including American ones. The UK has pursued a lighter-touch sectoral approach. China has its own suite of AI regulations focused heavily on content and recommendation systems. But none of these jurisdictions houses the companies building the most capable general-purpose models. OpenAI, Google DeepMind, Anthropic, Meta, and xAI are all American, which means U.S. policy choices — including the choice not to have explicit policy — effectively set the floor for global AI governance.

What the Trump administration has demonstrated, perhaps unintentionally, is that powerful governments cannot remain truly neutral toward transformative technologies. The economic stakes are too large, the national security implications too direct, and the leverage too obvious. The question is not whether Washington will shape the AI industry, but whether it will do so through transparent rules or through discretionary power that is harder to scrutinize, contest, or predict. For now, the answer is firmly the latter.