Recommendations for the U.S. AI Action Plan

Introduction

The United States (US) stands at a defining moment in the global contest for technological leadership. Its AI Action Plan (2025)^[1] sets an ambitious direction for AI development that is anchored in innovation, deregulation, and national resilience. The accelerating pace of technological and geoeconomic change, however, demands an even more precise and future-proof approach. As competing artificial intelligence ecosystems mature and the global economy increasingly demands dependable compute capacity, connectivity, and trusted digital infrastructure, the US can no longer rely solely on its domestic capability or current market dominance to ensure long-term leadership in the AI industry.

This report outlines a set of recommendations designed for the US’s AI Action Plan to extend and diversify the parameters of its strategic approach. The recommendations were drafted in response to the Request for Information in September 2025 issued by the US Office of Science and Technology Policy on how existing federal laws and regulations may impede AI development and adoption. Drawing on emerging trends in workforce disruption, infrastructure vulnerability, allied cooperation, public trust, and multilateral engagement, it argues that the next phase of US AI strategy must look beyond near-term competitiveness and toward long-term system resilience. By embedding anticipatory workforce planning, decentralising compute and energy networks, rebuilding data confidence, and deepening partnerships with both allies and swing states, the US can secure a durable advantage in an increasingly contested technological landscape. Strengthening these pillars will not only reinforce the US’s position in the AI century but shape global norms in a way that reflects its values and strategic interests.

Recommendations to Advance US Leadership in AI

The US AI Action Plan prioritises deregulation, innovation, and national resilience as a means to entrench the country’s leadership in the global AI economy. Yet, to ensure sustained dominance, the strategy must go further. Embedding foresight into workforce planning, securing resilient infrastructure, rebuilding public confidence, and coordinating with trusted allies will extend US influence globally. The following 10 recommendations outline a path to achieve that goal.

1. Prioritising partnerships over markets

An express purpose of the AI Action Plan is to export the US’s AI technology stack across countries to prevent and counter the spread of rival AI ecosystems. This policy approach has been consistent with the Framework for AI Diffusion enacted by the previous administration. Under the Trump administration, the US tech export policy is more pragmatic and open towards strategic partners, particularly with Gulf countries. However, the language used in the Plan falls short of considering countries within and adjacent to the US sphere of tech diffusion as equal partners, and instead relegates them to the status of mere “markets to capture”.^[2] 

The road to tech supremacy, however, is long and constantly evolving—it is nigh impossible to completely onshore the AI value chain. The threat of rival AI ecosystems, therefore, needs to be addressed through the creation of partnerships based on trust and shared goals. Towards this end, the broad approach outlined in the AI Action Plan must emphasise the necessity of mutually beneficial partnerships with nations across the world. A useful strategy can be to use the US-India TRUST (Transforming the Relationship Utilizing Strategic Technology) initiative^[3] as a template for bilateral agreements on technology cooperation with ‘swing’ countries in the Global South.

2. Building a future-proof workforce for the AI economy

Given the speed of development in AI, the particular AI skills in demand are subject to rapid change from year to year. For example, computer programmers, who have historically been highly sought after in the technology and software industries, have for the last couple of years seen around a 25-percent reduction in job opportunities as their tasks increasingly become automated by generative-AI tools.^[4] While the Plan correctly emphasises AI literacy and retraining, these measures risk being reactive rather than strategic. To safeguard the American workforce, the US should establish a National AI Workforce Foresight Council—a joint initiative between the Departments of Labor, Education, the National Science and Technology Council and the National Science Foundation—to model how automation and generative technologies will transform sectors over the next decade. Through close consultation with academics, economists, and industry innovators, the council would anticipate shifting skill demands before displacement occurs. This would ensure federal investments target emerging competencies, in addition to guarding against current retraining initiatives quickly becoming redundant.

3. Harmonising federal standards

The AI Action Plan emphasises the need for avoiding bureaucratic red-tape and onerous regulations that can hinder AI development. As it currently stands, however, the landscape of AI regulation in the US is a patchwork of laws at the state level. In 2024, state-level lawmakers in 45 states introduced 635 AI-related bills out of which 99 were passed.^[5] In 2025, around 260 AI-related bills were introduced by August, with 22 of them being passed.^[6] Meanwhile, since 2022, national strategies for AI regulation have been driven by a series of executive orders, putting them at the risk of discontinuity as White House administrations change.

Given that AI is a distributed technology, the absence of federal-level regulation will make the process of AI development and deployment unpredictable. For a company aiming to deploy an AI model nationally, this uncertainty creates a costly compliance matrix, discourages long-term investment and can create unnecessary internal data borders that hinder the flow of data, digital goods and services across the digital economy. The light-touch, industry-driven approach articulated in the AI Action Plan thus needs to be codified in federal regulation.  To address this, the next phase of the AI Action Plan should prioritise collaboration with Congress for passing baseline federal regulation aimed at harmonising standards nationally while also providing a long term and stable policy signal. 

4. Establishing a locally distributed data centre network 

The current US compute infrastructure landscape is dangerously concentrated in a few hyperscale providers clustered in specific regions. This bottleneck poses risks to economic resilience, national security and regional energy stability alike. In 2025, an incident in Virginia illustrated this vulnerability when roughly 60 data centres simultaneously disconnected from the grid, creating a sudden power surge and forcing operators to rapidly cut generation to prevent failures.^[7]  To mitigate these risks, the US should adopt a distributed compute strategy that incentivises the development of mid-sized, regionally based data centres powered by US-made semiconductors and locally sourced energy. Grants and tax credits should reward firms that decentralise infrastructure away from established centres, ensuring resilience against physical, cyber, or market shocks.

5. Diversifying the AI energy supply chain

The US AI Action Plan’s call to “Build, Baby, Build!” must be extended to the country’s energy foundations. AI data centres and semiconductor manufacturing facilities will continue to increase national energy demand exponentially. The US should prioritise a resilient mixed-energy grid by incorporating greater capacity and diversity from next-generation nuclear, geothermal and renewables produced domestically.^[8] China’s dominance in solar production, for instance, has strengthened its overall energy capacity, freeing resources to expand investment in energy-intensive AI systems. The country is also increasingly coordinating data centre energy demand with the construction of green electricity facilities.^[9] While the Trump administration has already identified increased fossil fuel extraction as a priority, developing additional domestic capacity in alternative scalable energy technologies would narrow the gap on that particular advantage while creating demand for the construction of new domestic manufacturing hubs for batteries and turbines. By expanding the sources of energy inputs that feed AI infrastructure, America further reduces its exposure to foreign supply chains.

6. Embedding coordination within a US-led AI bloc

AI leadership cannot be achieved in isolation, and should be pursued by adopting a leadership position within a strong alliance network. The US should extend the National AI Research Resource (NAIRR) pilot framework to include trusted allied research institutions under strict US-defined governance standards.^[10] Such coordination would expand access to frontier data and research while ensuring alignment with American norms on open innovation and free expression. Parallel consultations will harmonise regulatory sandboxes and end-use ethics codes to resolve inconsistencies. Ensuring interoperability of technologies among allies, in particular across defence and commerce, will reinforce US leadership in global rulemaking.

7. Reinstating AI safety leadership

The current AI Action Plan rightly prioritises infrastructure, talent, and investment but neglects safety frameworks that ensure US AI systems are trusted, interoperable, and—as a result—globally adopted. Recent proposals such as the AI LEAD Act, which seeks to make companies legally liable for unsafe AI products, reflect growing concern in Congress over accountability.^[11] Yet, overreliance on liability mechanisms could discourage firms from releasing advanced systems for fear of litigation. Establishing a strong, coordinated AI safety body would help mitigate that risk by ensuring products meet pre-deployment testing and certification standards before reaching the market. The Center for AI Standards and Innovation (CAISI), renamed from the AI Safety Institute in February 2025, should therefore be tasked not only with setting technical benchmarks but with making safety and security a driver of innovation. A reinstated National AI Safety and Security Directorate within the National Institute of Standards and Technology could coordinate these efforts, uniting government, academia and industry to ensure US-made AI becomes the global standard for secure, dependable, and high-performance systems.

8. Re-energising multilateral engagement

The AI Action Plan states the criticality for the US of countering Chinese influence in international bodies. This sentiment will remain theoretical without re-energising multilateral engagements and international partnerships. The US must aggressively fund expert participation in international standards-setting bodies like the International Organization for Standarization (ISO) and the International Electrotechnical Commission (IEC). The US must also use high-level forums like the G20 and the Organisation for Economic Co-operation and Development (OECD) to promote cooperation, best practices and interoperable policies. This can be achieved through formal coalitions with like-minded countries to build AI data commons that connect different ecosystems, as well as setting common-sense red-lines on AI development that can be viable alternatives to competing regulatory models.

9. Promoting reliability and stability

The US stack, which includes cloud infrastructure, foundation models, and open-source frameworks, is not just a product to be sold, but a foundational platform. Like any digital platform, it depends on widespread and deep adoption. Securing this global proliferation of the US AI ecosystem requires the prioritisation of stable and reliable access to American AI stacks, particularly for emerging markets.  Countries evaluating their national AI strategies need confidence that access to US tech will not suddenly be disrupted amidst shifting trade negotiations or domestic policy pivots. Nations may be forced to diversify supply chains and sell alternatives if the US is perceived as an unreliable partner. A first step towards signalling reliability can be mandating that new bilateral and multilateral trade agreements include clauses that protect access to defined commercial AI infrastructure and services from sudden interruption, with the exception of defined national security risks.

10. Strengthening public trust through transparent data governance and confidence-building initiatives

Public participation underpins AI success by ensuring that innovative models are built on the most comprehensive datasets possible. For example, voluntary genetic data sharing is foundational to US leadership in fields such as AI-enabled healthcare and genomics research, while access to high-quality mobility data is essential to providing accurate predictive models in the transportation industry. Yet, distrust of data collection, intensified by high-profile privacy and misuse scandals, threatens public cooperation in data sharing, as citizens fear loss of privacy, surveillance, or exploitation for commercial or political ends.^[12] Alongside robust safety standards to avoid data breaches, as outlined in the fifth recommendation, the US government should launch a National AI and Data Trust Initiative, led by the Office of Science and Technology Policy and the Department of Health and Human Services. This initiative would combine public education with strong enforcement of privacy guarantees, making it clear that data collection benefits the American people.

Conclusion

The recommendations outlined in this report point to a series of structural adjustments required to sustain US competitiveness in AI and supporting industries. Evidence across workforce capability, regulatory coherence, infrastructure resilience and international alignment, indicates that current arrangements must be improved to absorb future scaling and the likely pace of change associated with advanced AI systems.

A more anticipatory framework would help address these gaps. Integrating this report’s ten considerations into the next phase of the AI Action Plan will support US policymakers in ensuring that domestic AI infrastructure, governance mechanisms and international partnerships evolve in a manner that reinforce durable technological leadership over the long term. By coupling innovation with foresight, resilience and trust, the AI Action Plan can serve as a more stable framework for US technological leadership. In doing so, it would position the US to guide the evolution of global AI systems in a manner aligned with its strategic and governance priorities.

Siddharth Yadav is Fellow, Emerging Technologies, ORF Middle East.

Elizabeth Heyes is Junior Fellow, Emerging Technologies, ORF Middle East.

All views expressed in this publication are solely those of the authors, and do not represent the Observer Research Foundation, either in its entirety or its officials and personnel.

Endnotes

^[1] The White House, “America’s AI Action Plan: Winning the Race,” July 2025.

^[2] Sebastian Mallaby et al., “The Opportunities and Risks of Trump’s AI Action Plan,” Council on Foreign Relations, July 24, 2025.

^[3] The White House, “United States-India Joint Leaders’ Statement,” February 13, 2025.

^[4] Andrew Van Dam, “More than a quarter of computer-programming jobs just vanished. What happened?,” The Washington Post, March 14, 2025; Molly Kinder et al., “Generative AI, the American worker, and the future of work,” Brookings, October 10, 2024.

^[5] “2025 State AI Wave Building After 700 Bills in 2024,” Business Software Alliance.

^[6] “AI Legislation Across the US,” Retail Industry Leaders Association, September 8, 2025.

^[7] Tim McLaughlin, “Big Tech’s Data Center Boom Poses New Risk to US Grid Operators,” Reuters, March 20, 2025.

^[8] “What is US electricity generation by energy source?,” US Energy Information Administration, February 29, 2024.

^[9] “Country Rankings,” International Renewable Energy Agency, July 10, 2025; Nina Nurmamat, “China Hopes to Power AI Boom with Green Energy in New Data Centre Strategy,” South China Morning Post, June 5, 2025.

^[10] “National Artificial Intelligence Research Resource Pilot,” U.S. National Science Foundation.

^[11] Dick Durbin and Josh Hawley, “Aligning Incentives for Leadership, Excellence, and Advancement in Development (AI LEAD) Act,” United States Senate on the Judiciary.

^[12] Colleen McClain Park, Michelle Faverio, Monica Anderson and Eugenie, “How Americans View Data Privacy,” Pew Research Center, October 18, 2023.