America's AI Action Plan

America's AI Action Plan, subtitled "Winning the Race", is a federal AI strategy document issued by the White House in July 2025 (Source: whitehouse.gov). Its introduction is signed by Michael J. Kratsios, Assistant to the President for Science and Technology; David O. Sacks, Special Advisor for AI and Crypto; and Marco A. Rubio, Assistant to the President for National Security Affairs. The plan was produced under Executive Order 14179, "Removing Barriers to American Leadership in Artificial Intelligence," which directed the creation of an AI Action Plan and was signed after the Trump administration rescinded Biden's Executive Order 14110. It is organized around three pillars (innovation, infrastructure, and international diplomacy and security) and frames AI policy primarily through global competition and American leadership rather than risk management.

Framing and cross-cutting principles

The document opens with a quotation from Donald J. Trump describing AI as part of "a new frontier of scientific discovery" and stating that "it is a national security imperative for the United States to achieve and maintain unquestioned and unchallenged global technological dominance." The introduction states that the United States "is in a race to achieve global dominance in artificial intelligence," that "whoever has the largest AI ecosystem will set global AI standards and reap broad economic and military benefits," and that, as with the space race, the United States and its allies must win. It describes the potential of AI as "an industrial revolution, an information revolution, and a renaissance—all at once."

The introduction cites Vice President JD Vance's remark at the Paris AI Action Summit in February 2025 that restricting AI development with onerous regulation "would not only unfairly benefit incumbents… it would mean paralyzing one of the most promising technologies we have seen in generations," and states that President Trump rescinded the Biden Administration's actions on AI on his first day in office. On infrastructure it states that the administration will "continue to reject radical climate dogma and bureaucratic red tape" and summarizes the goal as "Build, Baby, Build!" It also states the aim of establishing American AI, "from our advanced semiconductors to our models to our applications," as the global gold standard on which allies build.

The plan names three principles that cut across all three pillars:

1. American workers are central to the administration's AI policy; the infrastructure buildout will create high-paying jobs, and AI will improve Americans' lives "by complementing their work—not replacing it."
2. AI systems "must be free from ideological bias and be designed to pursue objective truth rather than social engineering agendas when users seek factual information or analysis."
3. Advanced technologies must be protected from misuse or theft by malicious actors, with monitoring for "emerging and unforeseen risks from AI."

The introduction describes the plan as a set of policy goals "for near-term execution by the Federal government" and as a "roadmap to victory" in the AI race.

Pillar I: Accelerate AI innovation

The pillar states that America must have the most powerful AI systems in the world and lead in their application, and that the federal government should create the conditions for private-sector-led innovation.

Deregulation and state regulation

Under "Remove Red Tape and Onerous Regulation," the plan states that AI is "far too important to smother in bureaucracy at this early stage, whether at the state or Federal level." It says the federal government should not direct AI-related federal funding "toward states with burdensome AI regulations that waste these funds," while also not interfering with "states' rights to pass prudent laws that are not unduly restrictive to innovation." Recommended actions:

• An OSTP-led Request for Information from businesses and the public on federal regulations that hinder AI innovation and adoption.
• An OMB-led effort, consistent with Executive Order 14192 of January 31, 2025 ("Unleashing Prosperity Through Deregulation"), to identify, revise, or repeal regulations that unnecessarily hinder AI development or deployment.
• OMB work with agencies that run AI-related discretionary funding programs to consider a state's AI regulatory climate in funding decisions and to limit funding where a state's AI regulatory regime may hinder the effectiveness of the award.
• An FCC evaluation of whether state AI regulations interfere with the agency's obligations under the Communications Act of 1934.
• Review of all FTC investigations begun under the previous administration to ensure they do not advance theories of liability that unduly burden AI innovation, and review of FTC final orders, consent decrees, and injunctions to modify or set aside any that do.

Free speech and "American values" in frontier AI

Recommended actions under "Ensure that Frontier AI Protects Free Speech and American Values":

• The Department of Commerce, through NIST, revises the NIST AI Risk Management Framework to eliminate references to misinformation, Diversity, Equity, and Inclusion, and climate change.
• Federal procurement guidelines are updated so that the government contracts only with frontier LLM developers whose systems are "objective and free from top-down ideological bias."
• Commerce, through NIST's Center for AI Standards and Innovation (CAISI), researches and publishes evaluations of frontier models from the PRC "for alignment with CCP talking points and censorship."

Open-source and open-weight AI

Under the heading "Encourage Open-Source and Open-Weight AI," the plan gives three reasons for supporting such models: they "have unique value for innovation because startups can use them flexibly without being dependent on a closed model provider"; they "benefit commercial and government adoption of AI because many businesses and governments have sensitive data that they cannot send to closed model vendors"; and they "are essential for academic research." Recommended actions (see Open-Weight Frontier Models):

• Improve access to large-scale computing for startups and academics by improving the financial market for compute, in collaboration with industry, NIST, OSTP, and the NSF's National AI Research Resource (NAIRR) pilot.
• Partner with technology companies to increase the research community's access to private-sector computing, models, data, and software as part of the NAIRR pilot.
• Publish a new National AI R&D Strategic Plan, led by OSTP.
• Have Commerce, through NTIA, convene stakeholders "to help drive adoption of open-source and open-weight models by small and medium-sized businesses."

Adoption

The plan states that the bottleneck to realizing AI's potential "is not necessarily the availability of models, tools, or applications" but "the limited and slow adoption of AI, particularly within large, established organizations," and names healthcare as a sector that is especially slow to adopt. Recommended actions are regulatory sandboxes or AI Centers of Excellence around the country; domain-specific efforts led by NIST, for example in healthcare, energy, and agriculture, to accelerate national standards for AI systems; and regularly updated joint DOD–intelligence community assessments, led by DOD with ODNI, of the comparative level of AI adoption by the United States, its competitors, and its adversaries.

Workers

The plan describes a "worker-first AI agenda." It states that AI can create "more pathways to economic opportunity for American workers" but "will also transform how work gets done across all industries and occupations, demanding a serious workforce response." Recommended actions:

• DOL, ED, NSF, and Commerce prioritize AI skill development as a core objective of education and workforce funding streams.
• The Treasury issues guidance that many AI literacy and skill-development programs may qualify as eligible educational assistance under Section 132 of the Internal Revenue Code.
• BLS and Commerce, through the Census Bureau and BEA, study AI's impact on the labor market.
• An AI Workforce Research Hub is established under the Department of Labor (DOL).
• DOL uses available discretionary funding for rapid retraining of individuals affected by "AI-related job displacement."

The plan thus pairs its statement that AI will "complement, not replace" workers with retraining programs for displaced workers (see AI Labor Disruption).

Manufacturing, science and data

Further goals in the pillar cover next-generation manufacturing of technologies such as autonomous drones, self-driving cars, and robotics by the United States and trusted allies; AI-enabled science, noting that AI systems can already generate models of protein structures and novel materials and that general-purpose models show promise in formulating hypotheses and designing experiments; and world-class scientific datasets, which the plan calls "a national strategic asset," stating that other countries, including adversaries, have "raced ahead" in amassing scientific data and that the United States must lead in creating the largest and highest-quality AI-ready scientific datasets. A section on advancing the science of AI states that future breakthroughs may be as consequential as LLMs and generative AI were.

Interpretability, control and evaluations

The plan states that "the inner workings of frontier AI systems are poorly understood," which can make advanced AI hard to use in defense, national security, or other applications "where lives are at stake." It recommends a DARPA-led technology development program, with CAISI and NSF, to advance AI interpretability, AI control systems, and adversarial robustness, and making these a priority in the forthcoming National AI R&D Strategic Plan. It also calls for building an AI evaluations ecosystem, describing evaluations as how the industry assesses the performance and reliability of AI systems and as a tool for measuring reliability in regulated industries.

Government and defense adoption

For federal adoption, the plan recommends formalizing the Chief Artificial Intelligence Officer Council (CAIOC) (see Chief AI Officer (CAIO)); a talent-exchange program for rapid details of federal staff with specialized AI skills between agencies; an AI procurement toolbox managed by GSA; and a mandate that all federal employees whose work could benefit from frontier language models have access to them. For the Department of Defense (DOD), which the plan says AI could transform in both warfighting and back-office operations, it proposes an AI and Autonomous Systems Virtual Proving Ground and priority access to cloud computing for DOD during a national emergency.

Protecting innovations and synthetic media

The pillar also calls for government–industry cooperation to balance the dissemination of AI technologies with national security concerns, and addresses synthetic media in the legal system, stating that AI-generated media "may present novel challenges" and that "fake evidence could be used to attempt to deny justice to both plaintiffs and defendants."

Pillar II: Build American AI infrastructure

The pillar states that AI "is the first digital service in modern life that challenges America to build vastly greater energy generation than we have today," and that America's path to AI dominance depends on reversing that trend (see AI Data Centers).

• Permitting. The plan states that the environmental permitting system and other regulations "make it almost impossible to build this infrastructure with the speed that is required," and calls for streamlined permitting, including streamlined NEPA review, for data centers, semiconductor manufacturing facilities, and energy infrastructure, and for opening federal lands to data center construction, while guaranteeing security.
• Grid. The grid strategy is to stabilize the grid of today, optimize existing grid resources, and prioritize interconnection of reliable, dispatchable power sources while embracing new generation sources: enhanced geothermal, nuclear fission, and nuclear fusion.
• Semiconductors. Semiconductor manufacturing is to return to U.S. soil through a revamped CHIPS Program Office, with "extraneous policy requirements" removed from the program.
• High-security data centers. Because AI systems are well suited to processing raw intelligence data, data centers for military and intelligence community use must resist attacks "by the most determined and capable nation-state actors."
• Infrastructure workforce. A national initiative targets the trades needed to build and operate AI infrastructure, including electricians and advanced HVAC technicians.
• Cybersecurity. The plan states that as AI systems advance in coding and software engineering, "their utility as tools of both cyber offense and defense will expand." It calls for bolstering critical-infrastructure cybersecurity, including information sharing through an AI Information Sharing and Analysis Center (AI-ISAC), and for promoting secure-by-design AI, noting that AI systems are susceptible to adversarial inputs such as data poisoning and privacy attacks (see AI and Cybersecurity).
• Incident response. It calls for mature federal capacity for AI incident response so that, if systems fail, impacts on critical services or infrastructure are minimized.

Pillar III: International AI diplomacy and security

The pillar states that America must "drive adoption of American AI systems, computing hardware, and standards throughout the world."

• Exports. The United States is to export "its full AI technology stack—hardware, models, software, applications, and standards—to all countries willing to join America's AI alliance" (see EO 14320).
• International governance. The plan states that many international efforts "have advocated for burdensome regulations, vague 'codes of conduct' that promote cultural agendas that do not align with American values, or have been influenced by Chinese companies attempting to shape standards for facial recognition and surveillance," and calls for countering Chinese influence in international governance bodies.
• Export controls. It calls for strengthening enforcement of AI compute export controls, describing denial of advanced compute to adversaries as "a matter of both geostrategic competition and national security"; for plugging loopholes in semiconductor manufacturing export controls, noting that the United States and its allies hold "near-monopolies on many critical components and processes"; and for aligning protection measures globally by encouraging partners and allies to follow U.S. controls (see export controls).
• National security evaluations. The plan states that the most powerful AI systems "may pose novel national security risks in the near future in areas such as cyberattacks and the development of CBRNE weapons," and tasks CAISI at the Department of Commerce with evaluating frontier models for these risks.
• Biosecurity. It states that AI could "create new pathways for malicious actors to synthesize harmful pathogens and other biomolecules," and proposes "a multi-tiered approach designed to screen for malicious actors," with new screening tools and infrastructure, including required nucleic acid synthesis screening for federally funded research (see AI Biosecurity).

Analysis and relationships

The plan presents AI governance primarily as a matter of competition and leadership rather than risk management. Rather than directly preempting state AI laws, as EO 14365 attempted, it conditions federal funding on a state's regulatory climate, a mechanism the document positions as less direct, and pairs it with the FCC review of state AI regulations. The directed revision of the NIST AI RMF would remove topics (misinformation, DEI, climate) from a voluntary framework that many state laws reference, with potential effects on the standing of NIST AI RMF 1.0. The plan's stated position that AI will "complement, not replace" workers sits alongside its retraining provisions for AI-related job displacement.

The biosecurity provisions on nucleic acid synthesis screening are consistent with positions advocated by Dario Amodei, while the plan's overall posture contrasts with the graduated regulatory approach he has argued for. The DARPA interpretability and control program aligns with research priorities pursued at Anthropic. As a comprehensive statement of the Trump administration's AI policy, the document complements EO 14365, which focused on preemption. Its full-stack export strategy operationalizes the export controls debate; its workforce provisions engage the AI Labor Disruption debate from a more optimistic posture; and its framing on countering Chinese influence connects to AI and Authoritarianism and the Eight Worlds Framework.

Relationships

• supports: Open-Weight Frontier Models — the plan's Pillar I states an affirmative federal position in favour of open-source and open-weight models and directs NTIA-led promotion of their adoption.
• depends-on: EO 14179 — the executive order that directed the creation of the Action Plan.
• related: EO 14110 — the Biden executive order whose rescission the plan cites, describing it as having "foreshadowed an onerous regulatory regime."
• related: America's AI Action Plan — structured legislation-tracker page for the same document.
• related: EO 14365
• related: EO 14320 — full-stack AI export program.
• related: NIST AI RMF 1.0 — directed revision to remove references to misinformation, DEI, and climate change.
• related: CAISI — tasked with PRC-model evaluations and national-security evaluations of frontier models.
• related: NAIRR
• related: Dario Amodei
• related: export controls
• related: AI Labor Disruption
• related: AI and Authoritarianism
• related: Eight Worlds Framework
• related: AI Biosecurity
• related: AI Data Centers

Provenance

Issued by the White House (Kratsios, Sacks, Rubio), July 2025. Source file: Raw Sources/Americas AI Action Plan.md, captured from the White House PDF at whitehouse.gov. The raw capture reproduces the introduction and the recommended policy actions for the deregulation, free-speech, open-weight, adoption, workforce, interpretability, and government-adoption sections in full, but abridges most other sections to their opening paragraphs. Specific measures described above that are not in the capture (the Virtual Proving Ground, DOD priority cloud access, NEPA and federal-lands permitting, the AI-ISAC, the CHIPS "extraneous policy requirements" language, and nucleic acid synthesis screening for federally funded research) come from the full published plan. Confidence is medium: the page summarizes a single primary document.