How Trump's New AI Strategy Shifts Away from Safety Rules Toward Industry Growth

The Trump administration has made a sharp turn in how the federal government handles artificial intelligence. Instead of the safety-first approach of the previous administration, the new strategy leans heavily on letting companies innovate with fewer restrictions, expanding AI exports to allies, and partnering with industry leaders to shape policy.

The most significant move came on January 20, 2025, when President Trump cancelled Executive Order 14110 — the Biden administration's "Safe, Secure, and Trustworthy Artificial Intelligence" directive from October 2023. That order had required companies to conduct safety tests on large language models (the AI systems that power tools like ChatGPT), report their findings to the government, and assess potential risks before deployment.

A New Direction: Less Oversight, More Growth

Three days into his presidency, Trump signed Executive Order 14179, "Removing Barriers to American Leadership in Artificial Intelligence," on January 23, 2025. The title signals the shift: rather than focusing on potential risks, the focus became clearing away regulations that might slow down development.

Two more AI-related executive orders followed. One focused on AI education for young Americans (April 23, 2025), and another aimed to harness AI for pediatric cancer research with $50 million in additional funding (September 30, 2025). Instead of broad safety rules for all AI, this approach targets specific, high-impact problems.

In July 2025, the Trump administration released a blueprint for AI policy that included plans to ease environmental regulations and expand AI exports to allied nations. Under this framework, the Commerce and State departments will work directly with technology companies to package AI hardware, software, and standards for allied countries — essentially treating AI as a strategic asset for international partnerships.

Changes Inside Government and a Focus on Global Competition

The National Institute of Standards and Technology, which oversees federal technical standards, renamed its AI Safety Institute Consortium to simply the "NIST Artificial Intelligence Consortium." The name change reflects a broader shift: less emphasis on "safety" as a standalone concern, more on general AI advancement.

Michael Kratsios, who shaped AI policy during the first Trump administration, returned to advise on the current strategy. The administration's instinct is that American competitiveness depends on beating rivals internationally, and that happens by removing barriers and sharing technology with allies rather than holding it back.

This export strategy marks a real pivot. The previous administration had tried to limit where advanced AI could go, worried about misuse by adversaries. The Trump approach: share with friends, restrict only actual adversaries, and let American companies compete globally.

What Stayed in Place

Not everything changed. Executive Order 13960, which guides how federal agencies themselves use AI internally, remains active. That order requires government departments to track which AI tools they use and build up expertise in deploying them responsibly.

The Department of Homeland Security completed its AI safety report for critical infrastructure — power grids, water systems, communications networks — as required under the now-rescinded Biden order. This work continues under older cybersecurity rules, so these guidelines remain in effect even as the broader AI strategy shifted.

Federal agencies still collect data on how AI is being designed, developed, and used. This information gathering could inform future rules, even though the current administration prefers to move away from restrictive regulation.

Why This Matters, and What Questions Remain

We have seen this pattern before. In the early days of the commercial internet, the Clinton administration chose light regulation to let innovation flourish. That approach worked to grow the internet rapidly, but later administrations struggled with problems — from privacy breaches to fraud — that emerged from that early permissive era. The question now is whether a similar dynamic will play out with AI.

The administration's bet is straightforward: American companies and markets will police themselves better than government rules can, and staying ahead globally is the greater priority than preventing potential harms in advance. For companies, this is good news in the short term — no more mandatory safety testing or government reporting requirements. It also opens doors to export markets that were restricted before.

The bigger question, worth flagging, is that technology policy now swings sharply with each election cycle. That kind of instability makes it harder for companies to plan long-term research and for the government to project stable leadership on global AI standards. The faster AI evolves, the more that regulatory whiplash could become a real problem — though that's a challenge that will likely outlast any single administration as this technology continues to reshape industry after industry across the world.