Trump Administration Shapes 'Shadow AI Policy' Through Executive Actions, Tech Companies Face Regulatory Uncertainty
N.R. Finch
The Trump administration is bypassing Congress to shape AI industry rules through export controls, voluntary reviews, and procurement guidelines. This means → tech companies face not deregulation but an unwritten, shifting rulebook with no formal standards.
What happened to the "hands-off" promise?
Trump entered office pledging less AI regulation, revoked several Biden-era requirements, and warned that strict rules would stifle innovation.
In practice, the administration has pivoted to case-by-case corporate interventions, voluntary compliance frameworks, and executive orders — all running in parallel.
In plain terms = the government says "we won't regulate," but it is regulating — just without going through formal rulemaking. This reflects the executive branch's reflex to fill the gap when Congress stalls.
Why can't Congress legislate?
The House has introduced a bipartisan AI safety bill, but with midterm elections approaching, Congress as a whole is gridlocked on AI legislation.
That vacuum hands the initiative to the executive branch — the White House is steadily claiming policy ground through administrative action.
This means → no national AI law is coming in the near term. Companies are left reading executive orders and voluntary frameworks to guess at the rules.
What does the "shadow policy" actually govern?
Four policy directions are already clear: pre-empting some state-level AI laws, addressing national-security risks from advanced models, integrating AI into federal procurement, and assessing the economic impact of major AI firms.
The core tools are export controls, voluntary testing frameworks, and procurement guidelines — not statutes, but they carry real operational force.
In plain terms = there is no law called "the AI Act," but if you want to sell a model, win a government contract, or export chips, every step has a gatekeeper.
What does Anthropic's predicament reveal?
Anthropic is negotiating with the government to lift export restrictions on its latest model; other major labs are studying how to comply with Trump's newest executive order, which establishes a voluntary government review framework for certain advanced models.
Separately, the General Services Administration (GSA) is considering rules requiring large language models to strengthen data security when handling government information — setting privacy and security standards for federal contracts.
This means → the industry's core anxiety is not "too much regulation" but opaque regulation — companies must navigate not just policy, but personal relationships and broader political dynamics.
What are countries fighting over at the G7?
At this week's G7 summit, several AI company CEOs discussed with Trump the possibility of establishing a global AI standards forum; OpenAI policy chief Chris Lehane described it as "a global forum for AI standards."
"Technological sovereignty" and reducing dependence on American AI companies emerged as key concerns — yet leaders also acknowledge they can neither ignore the world's most advanced AI models nor fully replace them with domestic alternatives.
This reflects a deeper reality: the U.S. controls the most advanced AI models, and its policy choices — even informal "shadow policies" — carry the greatest weight for the global AI industry.
Content is for reference only, not financial advice.