January 15, 2026
The Honorable Jay Obernolte
Chair
Subcommittee on Research and Technology
U.S. House of Representatives
Washington, DC 20515
The Honorable Haley Stevens
Ranking Member
Subcommittee on Research and Technology
U.S. House of Representatives
Washington, DC 20515
Re: Written Submission for the Record – “Advancing America’s AI Action Plan”
Dear Chair Obernolte and Ranking Member Stevens,
On behalf of National Taxpayers Union Foundation, I respectfully submit this letter for the record in connection with the Research and Technology Subcommittee hearing entitled “Advancing America’s AI Action Plan” on January 14, 2026.
This letter examines a set of implementation and institutional design issues raised by the Administration’s Action Plan that are directly relevant to the Committee’s oversight and legislative responsibilities.1 In particular, it addresses the role of Congress in establishing durable statutory foundations—through reauthorization of the National Artificial Intelligence Initiative Act (NAIIA)—for initiatives that currently rely heavily on executive coordination, including AI evaluations, regulatory sandboxes, and shared research infrastructure.
First, the Action Plan assumes institutional scaffolding that only Congress can establish.
Many of the Action Plan’s core initiatives—including expanded AI evaluations, shared federal AI research and evaluation infrastructure anchored in the National Artificial Intelligence Research Resource, and regulatory sandboxes—currently rely on executive coordination rather than statutory design. While coordination through the Office of Science and Technology Policy is meaningful, it is inherently fragile: OSTP lacks independent regulatory authority, operates with limited budgetary depth, and remains vulnerable to shifts in leadership and political priorities.
Reauthorization of the National Artificial Intelligence Initiative Act is the primary vehicle through which Congress can establish durable coordination mechanisms, reporting obligations, and funding authorities, ensuring that these initiatives do not remain provisional or subject to policy volatility.
Second, federal AI evaluation capacity is expanding faster than its governance framework.
The Action Plan places heavy emphasis on building federal capacity for AI evaluations across the National Institute of Standards and Technology (NIST), the Center for AI Standards and Innovation (CAISI), the Department of Energy, and the National Science Foundation. While that is directionally correct, the Action Plan is largely silent on the procedural and institutional boundaries within which these evaluations will operate. Experience in other regulatory domains suggests that evaluation regimes that scale without clear rules can become difficult to bound institutionally, complicate relationships with regulated actors, and generate avoidable legal uncertainty—particularly where evaluative outputs begin to carry material consequences.
NAIIA reauthorization provides an opportunity to clarify the purpose and scope of federal AI evaluations, draw clearer distinctions between research, assessment, and enforcement functions, and establish baseline transparency and correction mechanisms. These steps would strengthen federal evaluation capacity rather than constrain it.
A related institutional risk is the blurring of strategic, technical, and enforcement-adjacent functions across the federal AI ecosystem. Strategy-setting, technical evaluation, and regulatory enforcement serve different institutional purposes and operate under different statutory authorities and accountability frameworks. Clear separation among these functions is not a constraint on federal capacity, but a prerequisite for sustaining it. NAIIA reauthorization provides an opportunity to clarify these roles early, reducing the risk that technical programmes are drawn into de facto regulatory or enforcement roles.
Third, regulatory sandboxes require a durable statutory framework.
The Action Plan recognizes the potential value of regulatory sandboxes as a tool for regulatory experimentation but offers little guidance on how they would function in practice. In the absence of statutory authority and shared design principles, sandbox initiatives risk remaining limited in practical effect, being implemented inconsistently across agencies, and lacking the institutional durability needed to persist across administrations.
NAIIA reauthorization provides an opportunity to make experimentation governable by authorizing both single-sector and cross-agency sandboxes aligned with the existing federal regulatory architecture, where AI systems cut across sectoral boundaries; ensuring that any regulatory relief is narrowly tailored and explicitly time-limited; and requiring agencies to document how sandbox outcomes inform subsequent guidance, standards, or legislative recommendations.2
Finally, regulatory coherence requires federal capacity, not preemption alone.
The Administration is right to identify fragmentation in state-level AI regulation as a challenge. However, a durable national framework cannot be built by displacing state rules before Congress has established the federal standards meant to replace them. Where clear federal baselines are absent, preemption risks removing existing obligations without supplying a governing architecture capable of operating across state boundaries.
The cross-cutting elements of AI governance—such as risk assessments, documentation standards, and transparency obligations—implicate systems and markets that operate across state lines and are therefore most appropriately addressed through federal baselines. By contrast, sector-specific rules governing AI use in employment, housing, health care, insurance, and consumer protection rest within long-standing state authority unless Congress clearly provides otherwise.
A more coherent approach would therefore focus on building federal capacity for the interstate elements of AI governance, while allowing sector-specific state regulation to persist—absent clear congressional direction to the contrary. Federal standards and any associated preemption may proceed in parallel, but preemption cannot operate meaningfully in the absence of the federal architecture it is intended to uphold.3
The AI Action Plan articulates a set of strategic priorities for the U.S. government. Whether those priorities are translated into a durable national framework now depends less on executive coordination than on congressional design. For this Committee, reauthorization of the National Artificial Intelligence Initiative Act is the clearest vehicle for establishing the regulatory architecture that the Action Plan assumes but does not itself provide.
Sincerely,
Ryan Nabil
Director of Technology Policy and Senior Fellow
National Taxpayers Union Foundation
122 C St NW, Suite 700
Washington, DC 20001
cc: The Honorable Brian Babin, Chair, House Committee on Science, Space, and Technology
The Honorable Zoe Lofgren, Ranking Member, House Committee on Science, Space, and Technology
1 The White House, Winning the Race: America’s AI Action Plan (Washington, DC: Executive Office of the President, July 2025), https://www.whitehouse.gov/wp-content/uploads/2025/07/Americas-AI-Action-Plan.pdf.
2 Ryan Nabil, “Why the United States Needs Better-Designed AI Sandboxes,” National Taxpayers Union Foundation, October 15, 2025, https://www.ntu.org/foundation/detail/why-the-united-states-needs-better-designed-ai-sandboxes.
3 Ryan Nabil, “Three Issues with the Trump Administration’s Proposed Preemption of State AI Laws,” National Taxpayers Union (blog), December 12, 2025, https://www.ntu.org/publications/detail/three-issues-with-the-trump-administrations-proposed-preemption-of-state-ai-laws.