5 Headlines: America's AI Action Plan
By Sushant Kumar, and Purva Agrawal.
America’s AI Action Plan aims to position the U.S. as the global leader in AI. Earlier this year, when DeepSeek-R1—an open-source LLM—was launched from China, many watchers called it America’s “AI Sputnik moment”. DeepSeek-R1 became the most-liked model of all time on Hugging Face—an open-source AI platform and community focused on sharing, training, and deploying AI models and datasets. Researchers, tech companies, labs, and startups utilised open source LLMs to build a multitude of their own variants. Therefore, the influence of Chinese AI foundational models in the US, and globally, is growing on the back of high-performing open source models.
The backdrop of the AI Action plan is China’s growing relevance in AI research, especially with wide global uptake of China’s open source models. AI was built on the foundations of open research. Increasingly, the American AI ecosystem became more closed and commercial, few steps removed from the open innovation of the recent past.
America’s response to the cold-war era Sputnik moment (1957 Soviet satellite) was fierce competition and a race to land humans on the moon, eventually culminating in an American victory. America’s response to China’s emerging AI leadership, with the AI Action Plan, has few bright spots. However, these bright spots are eclipsed by the displays of domestic political wrangling, grand geopolitical aspirations, and the desire to drop regulation altogether.
The AI Action plan is built on three pillars—Innovation, Infrastructure, and International Diplomacy and Security.
We share 5 relevant headlines that emerge from our assessment of the plan. They represent a mix of ambition, opportunity, and red flags.
1. Open Source, Open Weights, and Access to Compute: Back to the Basics
The U.S.’s plan has emphasised support for open source and open weight AI models. This is a welcome move, losing the open source race may mean losing the AI race, given its potential of accelerating innovation.
Increasingly, AI models made in the U.S. were locked behind gated access. On the other hand, models from China (such as by DeepSeek) released open source models preferred by American and global researchers for further research and fine tuning. They could even make “make these models their own” by tinkering. This significance makes the OS models integral to the foundational layer of global AI.
Moreover, the US plan tasks agencies like NSF, NIST, and OSTP to collaborate with the industry to increase access to compute through the NAIRR Pilot, helping establish a healthy financial market for AI compute and democratising AI development and innovation. (National Artificial Intelligence Research Resource (NAIRR) Pilot is a two-year initiative launched in January 2024 by the U.S. National Science Foundation (NSF) for developing a shared, national research infrastructure supporting responsible AI discovery, innovation, and education across the United States)
2. Err on the Side of AI Adoption: Mind the Gap!
The policy focuses on an adoption-first (try-first) culture for AI and outlines plans to introduce regulatory sandboxes and AI Centres of Excellence for rapid testing and deployment of new AI tools. Adoption assessment is intended to keep an iterative approach through comparative study across private and public sectors, competitors, and adversaries.
The policy emphasises on adopting AI for agencies in the government, enabling talent exchange programs for tech transfer and capability sharing. Inter-agency coordination is intended to promote AI adoption across the federal government. Additionally, it proposes an AI procurement toolbox to streamline uniform AI system access for federal agencies.
This focus on adoption in the government, while emphasising deregulation, sets a worrying trend.
3. Reskilling Workers: Bringing a Knife to a Gun Fight?
Job displacement and upheavals in the jobs market are a reality in the post-AI world. It is wonderful to see a “workers-first” agenda with recognition of this critical constituent. However, the outlined steps fall short and offer little details about implementation and accountability across agencies.
The plan proposes the following:
- Building AI skills: Agencies have been tasked with integrating AI literacy and technical skills into education, apprenticeships, and career training. Moreover, it proposes tax incentives for employer-provided AI skill programs.
- Retraining displaced workers: Fund and pilot skilling programs to help workers affected by AI automation to transition into new roles.
- Labor market research hub: Establish a Department of Labour hub to track AI’s impact on jobs, wages, skills, and displacement, thereby providing data to guide policy.
- Infrastructure workforce development: Identify and train workers for critical roles in AI infrastructure (e.g., electricians, data centre operators, HVAC technicians) through industry-driven apprenticeships.
4. Make AI Great Again: The Deregulation
The Action Plan promises to align AI systems with American values, prioritising “free expression and objective truth over ideological agendas”. The overarching MAGA theme is complemented with efforts on removing regulatory hurdles for AI Adoption – outlining a vision of AI with limited brakes.
There is focus on tackling synthetic media and for legal tools to combat deepfakes and misinformation. The plan emphasises AI evaluation systems, with guidelines to help agencies assess AI tools.
Notable steps include:
- Deregulation: Federal agencies have been directed to review existing AI regulations and cut back on any that hinder AI development or deployment. The plan seeks to discourage restrictive state AI laws through funding incentives.
- Regulatory sandboxes and standards development: The plan proposes AI regulatory sandboxes and Centres of Excellence to allow businesses and researchers to test AI technologies.
- Federal procurement policy: Procurement as a lever to promote AI models and systems free from “top-down ideological bias”.
- Coordination and enforcement: The White House Office of Management and Budget (OMB) will ensure federal resources do not support states with regulatory regimes deemed inconsistent with federal AI innovation goals.
5. AI Geopolitics: Overplaying the Hand?
The plan outlines strengthened export controls on sensitive AI technologies, including semiconductor manufacturing subsystems and advanced AI compute. These steps could spark a tech cold war that boxes everyone in—other countries will respond with their own restrictions similar to the ones imposed on critical minerals (magnets for batteries) by China recently. In a future that emphasises open innovation, these moves are counter productive and adversarial.
The Plan promotes active U.S. participation in global AI governance through international bodies such as the United Nations, OECD, G7, G20, and others. It seeks to shape international AI standards and policies that reflect American values.
Moreover, the plan outlines exporting the entire AI tech stack—hardware, software, models, and standards—to willing allies to foster an AI alliance. Further priorities include tech-diplomacy, military adoption of AI, and cyber security preparedness for AI infrastructure.
In Summary
America’s AI Action Plan’s focus on open innovation, workforce empowerment, and global leadership are steps in the right direction. However, regulatory approaches must be mature and geopolitical moves need strategic finesse. As AI transforms work and workforce, there is a need to establish global leadership that enables an equitable transition.
AI demands an inspired global leadership and not narrow domestic political walls. Open innovation coupled with the right values, safeguards, and access to resources (compute, data, talent) is the right path for America and any country that is interested in setting the course of humanity’s future of AI.