Unpacking the US AI Action Plan

On July 30, 2025, Gregory C. Allen, Senior Adviser at CSIS Wadhwani AI Center, sat down with Michael Kratsios, Director of the White House Office of Science and Technology Policy, to unpack the details of the new U.S. AI Action Plan.

The discussion opened with reflections on the Trump administration’s immediate efforts to overhaul existing AI regulations. Upon entering office, the administration revoked the previous executive order on AI, which had caused concern across the AI community for its perceived overreach and regulatory ambiguity. In its place, the new administration launched a comprehensive AI Action Plan prioritizing national leadership in AI as a matter of economic and national security.

Public engagement also played a major role in shaping this plan. Through an open call for input, the administration received over 10,000 responses from a diverse array of stakeholders, including AI companies, Hollywood figures, financial institutions, and civil society.

The plan is centered on three pillars:

1. Innovation leadership: Creating a supportive regulatory environment for AI innovation in the U.S., ensuring certainty for startups and small businesses, and boosting R&D investments.
2. Infrastructure development: Expanding datacenter capacity and power generation to meet AI’s growing demands, while addressing workforce shortages in several areas, including traditional professions such as electricians (identified as a challenging role to fill in the Stargate construction site).
3. Global AI tech stack dominance: Promoting widespread adoption of the US AI tech stack (including chips, models, and applications) while activating US development finance institutions to support international exports.

Kratsios emphasized the strategic nature of bundling the US AI stack as a turnkey solution, contrasting it with China’s full-stack approach exemplified by Huawei. He described the administration’s focus on ensuring global deployment of American AI tools while simultaneously tightening enforcement of export controls on high-end chips and semiconductor manufacturing equipment, especially regarding access by China.

The conversation also covered:

• Export control strategy: Maintaining strict controls on high-end semiconductors while increasing enforcement, including leveraging intelligence capabilities.
• State-level AI regulation: Concerns over a regulatory patchwork; the administration seeks to preempt state regulations through federal grants and FCC levers, aiming for uniform national rules.
• Engagement with international AI regulation: The US government opposes the European Union’s AI Act and its horizontal, one-size-fits-all approach, instead promoting sector-specific, risk-based regulation managed by domain-specific agencies.
• The future of CAISI (Center for AI Standards and Innovation, formerly the AI Safety Institute): Refocused on standards and measurement science under NIST, supporting model evaluations with concrete metrics rather than vague safety concepts.
• Labor and education: Kratsios described efforts to train a new workforce to support AI infrastructure and integrate AI into K–12 education to prepare students for future jobs.

The conversation concluded with a discussion of metrics for success (e.g., permitting speed for datacenter construction), balancing chip exports with domestic AI needs, and the administration’s readiness to act on emerging bottlenecks. While no specific legislative agenda was announced, areas for Congressional involvement include AI-related R&D funding, state preemption, fair use laws for training data, and formalizing CAISI’s role.

Full video and transcript of the event is available here.