The game is afoot: America’s new AI doctrine

In July 2025 the White House released America’s AI Action Plan, a document that redefines how the United States intends to compete in the global artificial-intelligence race. Signed by President Donald J. Trump and authored by senior officials Michael J. Kratsios, David O. Sacks, and Marco Rubio, the plan replaces Joe Biden’s 2023 executive order on “Safe, Secure, and Trustworthy AI.” Where the earlier framework emphasised caution, ethics, and civil-rights protections, this new doctrine declares a return to speed, power, and economic nationalism. It reads less like a regulatory charter and more like a wartime industrial strategy, one that treats algorithms as assets of national defense and data centers as strategic infrastructure.

At its core, the plan rests on three pillars: Accelerate AI Innovation, Build American AI Infrastructure, and Lead in International AI Diplomacy and Security. Each pillar translates political ideology into concrete directives designed to remove barriers, unleash private-sector ingenuity, and confront the geopolitical challenge posed by China’s state-driven technology model.

Pillar I – Accelerate AI Innovation

The first pillar dismantles the dense web of oversight introduced under Biden’s Executive Order 14110. It begins with a categorical instruction to eliminate “red tape and onerous regulation.” The federal government, the plan insists, must not attempt to govern artificial intelligence as a discrete technology class. Instead, agencies are ordered to identify and rescind rules that impede innovation. Funding will be withheld from states imposing “burdensome AI regulations,” a direct rebuke to the patchwork of state-level privacy and algorithmic-bias laws emerging in California, New York, and elsewhere.

What replaces regulation is a faith in competition and open markets. The plan introduces a Free Speech and American Values doctrine requiring that government-funded models be “objective and free from ideological bias.” The National Institute of Standards and Technology (NIST) is directed to remove every reference to “misinformation,” “diversity,” or “climate change” from its AI Risk Management Framework, arguing that such language politicises science. In a striking twist, NIST is also tasked with evaluating Chinese large-language models for alignment with Communist Party narratives—a fusion of technical benchmarking and ideological screening that underscores how the administration equates information integrity with national security.

Open-source AI becomes the plan’s intellectual and strategic centerpiece. The administration portrays open-weight models not as a privacy threat but as instruments of geopolitical leverage—a “geostrategic weapon” ensuring that American innovation sets the global standard. Through expansion of the National AI Research Resource, universities and startups are promised subsidised access to compute power and datasets previously concentrated in corporate hands. The message is unmistakable: where China builds state-controlled systems, America will dominate through open distribution.

Beyond deregulation, the plan sketches a network of AI Sandboxes in finance, healthcare, and agriculture. These zones exempt innovators from pre-approval requirements, allowing real-world experimentation under minimal supervision. In labour policy, two executive orders—Advancing AI Education for American Youth and Preparing Americans for High-Paying Skilled Trade Jobs of the Future—anchor a retraining push. Workers displaced by automation are to be “re-equipped,” not shielded. An AI Workforce Research Hub will measure productivity gains and redeploy skills accordingly. The underlying belief is that Americans, if unshackled by bureaucracy, can adapt as fast as the technologies transforming their lives.

Defense completes the first pillar. The Department of Defense is instructed to establish AI Proving Grounds for autonomous systems, ensure military access to high-performance compute during crises, and coordinate with the intelligence community on “AI net assessments” of foreign adversaries. Even the legal system is conscripted: the Department of Justice must integrate deepfake-detection standards into the Federal Rules of Evidence, extending NIST’s forensic program—aptly titled Guardians of Forensic Evidence—to protect elections and public trust.

Pillar II – Build American AI Infrastructure

If the first pillar frees the private sector, the second mobilises the state to build the physical backbone of an AI superpower. Here, energy policy, industrial policy, and environmental policy converge under the banner of national security.

The plan orders sweeping permitting reform to accelerate construction of semiconductor foundries, cloud facilities, and power plants. It expands exemptions under NEPA and related environmental statutes and even allows data-center development on federal lands. A new system dubbed PermitAI will automate environmental reviews, a technological solution to bureaucratic delay. The administration makes clear that “AI infrastructure” includes not only chips and servers but also pipelines, grid extensions, and generation capacity.

Energy strategy is equally explicit. The plan condemns what it calls “radical climate dogma” and commits to expanding natural gas, nuclear, and other “reliable sources.” Abundant electricity is portrayed as the foundation of digital sovereignty. Without it, the argument goes, America’s data centers—and by extension its intelligence apparatus—would depend on fragile foreign supply chains.

Cybersecurity threads through every section. All federal AI infrastructure must be free from “foreign adversary ICT hardware or software,” an unmistakable reference to Chinese-made components. The Department of Homeland Security and the Cybersecurity and Infrastructure Security Agency are to develop an AI-specific incident-response system capable of defending both civilian and military networks. Parallel to this technological mobilisation, a vocational program will train electricians, engineers, and technicians for the data-center boom the plan envisions. AI, in this narrative, is not a disembodied algorithm but a full-scale industrial revolution requiring concrete, steel, and skilled hands.

Pillar III – Lead in International AI Diplomacy and Security

The third pillar shifts from domestic policy to global power politics. It opens with a simple assertion: “The AI race is America’s to win.” To win it, Washington must export its technologies and its values while denying rivals access to critical inputs.

China occupies the center of this strategic horizon. The plan orders a tightening of export controls on advanced semiconductors, cloud-compute services, and model weights. Intelligence agencies are to track China’s frontier AI projects and assess their military relevance. The United States will also work through alliances to “synchronize export control frameworks,” preventing circumvention via intermediaries in Hong Kong or Singapore. In multilateral bodies such as the ISO and ITU, NIST and the Department of Commerce will campaign to enshrine American technical standards as global norms.

At the same time, the administration commits to promoting U.S. AI systems abroad through bilateral partnerships. Latin America, Europe, and Asia are identified as priority markets for U.S.-developed models and chips. The plan’s authors cast this as a twenty-first-century analogue to the Marshall Plan: exporting infrastructure and ideology simultaneously. Embedded within this diplomatic offensive is a biosecurity agenda directing new funding toward AI-enabled monitoring of pathogens and synthetic biology. In the age of autonomous science, the line between health defense and national defense dissolves.

Nowhere is the philosophical reversal from Biden’s era more visible than in the fate of the National Institute of Standards and Technology. Under Executive Order 14110, NIST was the guardian of “trustworthy AI,” responsible for fairness audits, bias mitigation, and civil-rights guidance. Under the 2025 plan, it becomes the technical backbone of a performance-first regime. The institute’s risk-management framework is rewritten to exclude ethical language; its mission redefined around robustness testing, model interpretability, and energy efficiency. Social trust gives way to operational reliability. Yet NIST retains one moral mandate: the fight against synthetic media. Its deepfake-detection program expands into a national standard, the only remaining bridge between ethics and enforcement in the new doctrine.

A Geopolitical Doctrine, Not a Regulation

Read as a whole, America’s AI Action Plan functions less as an industrial blueprint than as a declaration of geopolitical posture. It transforms AI policy into a contest of systems: open versus authoritarian, decentralized versus state-controlled. Where Europe writes rules, and China builds surveillance, the United States—this plan insists—will build faster, freer, and bigger. “Build, Baby, Build,” the text proclaims, rejecting environmental caution as an impediment to destiny.

The philosophical contrast with 2023 could not be sharper. Biden’s order sought to protect workers from automation; Trump’s seeks to turn workers into builders of automation. The former placed NIST at the intersection of ethics and governance; the latter reassigns it to the factory floor. Under the old regime, government acted as referee; under the new, it is the chief engineer.

This shift carries global implications. Allies who aligned with Washington’s earlier rhetoric on “trustworthy AI” must now reconcile themselves to a doctrine of raw competition. Companies that invested heavily in compliance frameworks may find those efforts politically obsolete, even as states dependent on U.S. cloud infrastructure discover that energy and security considerations now outweigh environmental pledges. For all its nationalistic tone, however, the plan reveals a deep continuity in American policy: an abiding belief that technological leadership is inseparable from geopolitical power.

The America’s AI Action Plan closes with the signatures of its architects beneath a triumphant claim that “the AI race is America’s to win.” Behind the bravado lies a coherent worldview. Artificial intelligence, in this doctrine, is not a moral puzzle to be solved but a strategic domain to be mastered. It is energy-hungry, capital-intensive, and inevitably political. In choosing deregulation over restraint, the administration wagers that freedom—and perhaps audacity—remain the ultimate American comparative advantage.

Whether this gamble pays off will determine more than the future of technology; it will shape the moral and industrial character of the twenty-first century. The United States has decided that in the age of machines, velocity is virtue. The rest of the world will now have to decide whether to keep up—or to hold the line.