
The Trump administration's newly released AI Action Plan aims to make good on the president's commitment to securing U.S. dominance in artificial intelligence. This includes accelerating innovation in AI by removing regulatory barriers to its development, streamlining the build-out of American AI infrastructure, and ensuring that the U.S. leads in AI internationally through standards-setting and support for American technology exports.
A giant leap in the right direction, the Action Plan is just the beginning of the administration's efforts to cement American leadership in AI. As the associated executive orders are signed, a litany of initiatives will kick off across the federal government, making critical the plan’s full and faithful implementation. This includes elements of the strategy that are still forthcoming, such as the National Artificial Intelligence R&D Strategic Plan—a key pillar of the administration’s commitment to advancing the science of AI itself.
In the bigger picture, however, the AI Action Plan is a breath of fresh air compared to either the Biden administration’s executive order on AI or the European Union’s sclerotic AI Act. While the plan proposes measures that indicate the administration takes the risks from AI seriously, particularly as they relate to national security, it doesn’t let the potential downsides of AI dominate its vision. Instead, it charts a proactive approach to unlocking AI’s enormous potential upside, mitigating risks where they exist not through sweeping, reactive regulation, but through smart investments in information sharing, oversight, and R&D.
Accelerating American AI
The AI Action Plan opens with a series of actions to remove bureaucratic red tape that could hinder AI development and undermine U.S. leadership. Importantly, it takes an open-ended approach, directing the Office of Science and Technology Policy and the Office of Management and Budget to lead efforts to “identify, revise, or repeal regulations, rules, memoranda, administrative orders, guidance documents, policy statements, and interagency agreements” that hinder AI development and adoption across all federal agencies.
As I argued in my Action Plan RFI submission and elsewhere, the biggest regulatory barriers to AI adoption are mostly not AI-specific, but rather buried in decades of accumulated regulatory detritus that codify the best practices of a bygone, pre-AI era. This is all the more true in highly regulated sectors such as health care, finance, and education—areas where AI coincidentally has many of its most promising near-term applications. Technologies as general purpose as AI thus necessitate a broader regulatory reset to take into account new capabilities and modes of doing business that prior generations of regulators simply failed to foresee.
The regulatory barriers to AI adoption are particularly severe in physical sectors. To that end, the Action Plan includes a robust strategy to support AI adoption for next-generation manufacturing. This will be key to ensuring the U.S. translates its leadership in AI software into real productive output. As I wrote in my RFI submission,
While the internet revolution was primarily won in the lightly-regulated realms of software and microelectronics, winning the AI race will require overcoming bottlenecks in the heavily-regulated physical world, from the energy generation needed to fuel data center growth, to AI’s countless applications in robotics, material science, biology and beyond. Indeed, as a side-effect of deindustrialization, the last 40 years have seen the United States deepen its comparative advantage in high value-added “knowledge” sectors, such as software development, entertainment, higher education, management, law and finance—the very sectors AI is most likely to disrupt and deflate in the coming years. Thus, without concerted efforts to reindustrialize and deregulate the physical economy—energy, transportation, manufacturing, construction, health care, etc.—America’s “innovation in bits” risks becoming China’s “innovation in atoms.”
The AI Action Plan’s support for American industry is further buttressed by its strategy for AI-enabled science. In line with my RFI recommendations, the Action Plan calls for federal scientific datasets to be made available for use in AI model training, along with investments in automated cloud-enabled labs and support for Focused-Research Organizations.
Lastly, the AI Action Plan further directs OMB to work with federal agencies to limit AI-specific discretionary funding to states with regulatory regimes inimical to AI development. Following the debate over the proposed moratorium on state-level AI regulation, this represents a reasonable incentive against states creating a regulatory patchwork, particularly insofar as the federal government has a prerogative to maximize the impact of its AI-specific funding.
Government Modernization
Beyond removing outdated regulations, the Action Plan takes proactive steps to enable AI adoption through regulatory sandboxes—“AI Centers of Excellence”—for researchers and businesses to rapidly deploy and test new AI tools. Notably, these sandboxes would be established under specific regulatory agencies, such as the Food and Drug Administration (FDA) and Securities and Exchange Commission, enabling the experimentation needed to inform novel, AI-native regulatory frameworks.
In a world where AI helps drive an explosion in the number of novel drug candidates, for example, it may not be sufficient to reform the FDA drug approval process on the margins. Instead, fundamental process reform will be needed to match the scale and pace of change—a case not of regulating less but regulating differently. A dynamic, “try-first” culture of experimentation in settings exempt from the regulatory status quo is thus essential to driving institutional learning and adaptation.
The AI Action Plan takes broader steps to accelerate the adoption of AI in government–an issue of critical importance for U.S. state capacity. As I noted in my RFI submission and efficiency agenda paper, the private-sector diffusion of powerful AI systems creates institutional risks that are “manageable if and only if the rate of AI diffusion within the public sector keeps pace with the private sector.” Among other things, the Action Plan thus aims to streamline adoption through an “AI procurement toolbox” and Advanced Technology Transfer and Capability Sharing Program to ensure that advanced AI capabilities can be rapidly adopted and transferred to new use cases between agencies.
In contrast with the Biden administration’s cumbersome and risk-averse approach to internal AI governance, the Action Plan mandates “that all Federal agencies ensure—to the maximum extent practicable—that all employees whose work could benefit from access to frontier language models have access to, and appropriate training for, such tools.” This in a sense flips the burden of proof on agencies’ Chief AI Officers, requiring them to justify why an AI system wasn’t made accessible rather than the other way around.
Free Speech and American Values
The AI Action Plan next takes aim at upholding American values and free speech. It does this by revising existing federal AI frameworks to remove references to Diversity, Equity, and Inclusion (DEI), climate change, and misinformation. This is an important step toward restoring public trust in AI standards, resetting the focus of documents such as the National Institute of Standards and Technology’s AI Risk Management Framework around actual risk management issues—not political or ideological agendas.
I’ve written before about how so-called “AI ethicists” cherry-pick examples of AI bias to demand that AI models be fine-tuned into being “woke,” making models even more biased in practice. Rather than impose a homogenous standard or micromanage how AI companies differentiate their products, the Action Plan focuses on standards of objectivity and impartiality for systems procured by the federal government. As the related executive order clarifies, this can be largely accomplished through basic transparency, e.g., disclosing the system prompts and fine-tuning methods used to align the model.
If there is any major omission in this section of the Action Plan, it is a strategy for dealing with AI’s intrinsic risks to Americans’ privacy and civil liberties. As I note in my RFI submission, multi-modal AI models dramatically lower the costs of government surveillance and censorship, demanding a systematic review of “the potential for AI-enabled mass surveillance through existing regulatory vectors” and strategies to safeguard against overreach.
Open Source and Tech Diplomacy
The AI Action Plan reverses the Biden administration’s hostility to open source and open weight AI, recognizing open weight models as both a boon for researchers and small businesses, and essential for use in critical infrastructure and sensitive government contexts. This parallels my recommendations to reform procurement to “promote adoption of open source AI and reduce compliance burdens that benefit Big Tech incumbents,” and to adopt a stance of “open source diplomacy” internationally.
Promoting the diffusing of the American AI technology stack internationally, including our models, hardware, and software ecosystems, is imperative to securing U.S. dominance in AI now and into the future. By using our first-mover advantage to win export markets abroad, we can ensure that countries around the world become integrated into U.S. technology ecosystems that both embed U.S. governance standards and make our technological lead more durable. The Action Plan does this by adopting policies that echo the recommendations made by my colleagues in “The Closing Window to Win.” In particular, the Plan calls for the creation of “full-stack AI export packages” developed in collaboration with industry consortia, which can then be diffused abroad with the backing of the full array of U.S. federal export promotion mechanisms.
Combined with the AI Action Plan’s initiatives to “Counter Chinese Influence in International Governance Bodies,” the Trump administration is treating the AI race as not merely technological but also a race to defend and diffuse American values. Given the alternative—a 21st century defined by the values of the Chinese Communist Party—we must make every effort to capture markets worldwide while containing the CCP’s access to our most advanced technology.
To that end, the Action Plan includes promising recommendations on how to bolster U.S. export controls on advanced AI chips and semiconductor manufacturing equipment. This includes a nod to location verification mechanisms, an approach I’ve advocated for as a scalable solution to China’s chip smuggling efforts.
Energy, Infrastructure, and Labor
The AI Action Plan takes decisive measures to accelerate investments in domestic AI infrastructure and energy resources. As discussed in my RFI submission, China’s much weaker energy and transmission bottlenecks leaves us vulnerable to being leap-frogged should they catch up on AI hardware. Streamlining permitting for U.S. data centers, transmission infrastructure, and baseload power generation is thus critical to maintaining our lead.
From expedited environmental permitting for data center and energy projects, to a framework for unlocking federal lands in regions that can support large AI clusters, the Action Plan reflects many of the recommendations developed by FAI’s director of infrastructure policy, Thomas Hochman. In turn, the administration is activating every tool available to shore-up grid resilience and enable the all-of-the-above energy build-out needed to match AI’s unprecedented projected demand.
Additionally, the Action Plan calls for activating federal workforce development programs to help close talent gaps in the skilled trades, from HVAC to construction. Paired with pilot programs to better track AI’s labor market impact and explore novel approaches to worker retraining, the Action Plan thus aligns Vice President Vance’s call for a “worker-first” approach to AI with the needs of the nation.
Frontier Model Safety and Security
The capability of frontier AI systems is improving rapidly, but there is still enormous uncertainty about both the types of risks AI poses and when those risks will first manifest. Nevertheless, the emergence of extremely powerful, autonomous AI systems is plausible within the next 18 months, posing immediate risks in areas such as cyber- and biosecurity. We shouldn't let skepticism over the most alarmist scenarios distract from the real challenges ahead of us, from labor market disruptions to the many bona fide risks future AI systems present to U.S. national security.
Fortunately, the AI Action Plan takes these imminent risks to heart, proposing a number of pragmatic steps to strengthen oversight of frontier AI companies and mitigate worst-case scenarios. This includes reaffirming the Center for AI Standards and Innovation’s role in evaluating frontier AI systems for national security risks, biosecurity investments to close pathways for malicious actors to synthesize harmful pathogens, support for research into AI control and interpretability, and a new AI Information Sharing and Analysis Center led by the Department of Homeland Security to enable prompt incident reporting and communications between the leading AI companies and the U.S. government.
Basic monitoring and oversight of top AI labs has long been my primary recommendation as a prerequisite for frontier AI security, and has only become more important as we inch closer to transformative AI capabilities. The U.S. government should not be caught unawares by AI developments with direct implications for things like cybersecurity. To date, the top AI labs have been good about reporting incidents and sharing information with national security stakeholders on a voluntary basis and through partnerships, but their capacity to omit or delay the sharing of relevant information shouldn't even be an option.
Winning the AI Future
In conclusion, the Trump administration’s AI Action Plan goes beyond a hands-off approach to AI in favor of proactively removing barriers to AI development; harnessing AI to drive American science and innovation; and embracing an export-oriented strategy to ensure that America's AI technology stack secures dominance in markets around the world.
While there is much to applaud in the Action Plan, it is only the start. AI promises to be the most transformative technology in human history, but with dramatic societal transitions comes a need for governments to be nimble and adaptable. Securing America’s dominance in AI is thus not merely important for economic competitiveness, but critical for U.S. national security and, ultimately, the preservation of the American way of life.