Trump Administration Announces New Direction for U.S. AI Policy

On July 23, 2025, the Executive Office of the President released Winning the Race: America’s AI Action Plan (“AI Action Plan” or the “Plan”), a new national strategy for the Federal government to establish U.S. leadership in artificial intelligence (“AI”).^[1]  It outlines over 90 recommended policy actions for federal agencies falling under three pillars.  The Plan marks a major policy shift, replacing the rescinded Biden Administration AI order^[2] and introducing new guiding principles, including mandating a review of AI systems guidelines to be “unbiased,” and providing recommended policy actions suggesting a comprehensive AI regulatory framework.

Background

In early 2025, the White House began to redirect U.S. AI policy when President Trump issued Executive Order 14179, Removing Barriers to American Leadership in AI, which rescinded Biden’s AI Executive Order and called for the establishment of a new plan for artificial intelligence.^[3]  The administration developed and published the AI Action Plan to conform to those mandates.^[4]  The Plan seeks immediate and sweeping changes to Biden-era AI governance frameworks, which expressed concern about potential societal, civil rights, and national security risks associated with AI.  The new plan pursues the objective of a “try first” approach towards AI, calling on federal agencies to remove rules that could impede AI innovation and to use existing authority to accelerate AI growth across federal agencies and various industries.^[5]  The AI Action Plan also cautions that federal funding should not be provided to states with “burdensome AI regulations” that may hinder innovation.^[6]

Key Implications from Each Pillar

The Plan is organized around three key pillars: (1) Accelerate AI Innovation–removing regulatory hurdles, promoting open-source and open-weight AI, and rapidly integrating AI in various sectors; (2) Build American AI Infrastructure–fast-tracking data center permits, chip fabrication, power grid upgrades, and workforce development; and (3) Lead in International AI Diplomacy and Security–securing U.S. supply chains, tightening export controls on adversaries, enhancing enforcement, and shaping global AI standards alongside allies.^[7]

Notably, the Plan relies on executive actions and agency policy changes rather than new laws or a comprehensive AI framework, and it exists alongside a growing patchwork of AI-specific state and international regulations and laws (like Colorado’s Artificial Intelligence Act^[8] and the European Union’s AI Act^[9]).  This dual context is important: the federal Plan seeks an act-first approach, while organizations may still need to heed more cautious state rules and global standards in practice.

Pillar I:  Accelerate AI Innovation 

The White House’s AI Action Plan signals an aggressive deregulatory policy for AI.  It introduces no new regulations or statutes and creates federal policies that direct agencies to eliminate “onerous” constraints on AI development.  These mandates ostensibly shift the onus on companies to self-govern, while also requiring AI systems to be free from “ideological bias.”

The Action Plan seeks to achieve innovation through deregulation with its recommended policy actions, including having the Office of Management and Budget (“OMB”) coordinate with federal agencies with discretionary AI funding authority to consider a state’s AI regulatory scheme when funding state programs.^[10]  This federal/state dynamic is something to watch for as states battle to attract both federal and private investment in state infrastructure to meet the needs of their AI programs.

Mandate for Ideologically Neutral AI:  The Plan directs agencies involved in procurement to create AI policies, systems, and programs that are ideologically neutral.  AI systems, especially those used by government, are to be “objective” and free from political or ideological bias.  In practice, the Plan proposes updated federal procurement guidelines to require that the government contract only with frontier large language model (LLM) developers with “Unbiased AI Principles” and that their models do not exhibit partisan bias or censor lawful speech.

Pillars 2 & 3:  Build American AI Infrastructure and Lead in International AI Diplomacy and Security

Pillars 2 and 3 of the AI Plan highlight the need for an expansion of American energy capacity to fuel the development of an AI infrastructure, including building new chip manufacturing factories, data centers, and tapping new sources of energy.  Recommended actions include reforming environmental policy related to permits for new data centers, among others.  Infrastructure expansion includes workforce training, restoration of American semiconductor manufacturing, and building high-security data centers for the military and intelligence communities. 

According to the Plan, global expansion of American AI systems can be achieved through the export of a technology stack of computing hardware, models, applications, and standards to allies and international partners.  The plan also urges the pursuit of creative means of AI compute export control enforcement to promote U.S. AI objectives around the world.  The goal is for American AI to be the standard bearer, which would create partnerships with allies to prevent their reliance on rivals.

Key Dates

While the Plan itself does not include specific deadlines for regulatory action, it does correspond to three Executive Orders that establish timeframes:

Executive Order “Preventing Woke AI in the Federal Government” (July 23, 2025) instructs the OMB and Administrators for Federal Procurement Policy and General Services, and the Director of the Office of Science and Technology Policy (OSTP) to issue guidance within 120 days to procure only LLMs developed using “Unbiased AI Principles” (which the administration terms “Truth-seeking” and “Ideological Neutrality”).^[11]

Executive Order “Accelerating Federal Permitting of Data Center Infrastructure” (July 23, 2025) directs the EPA within 180 days to develop guidance to expedite environmental reviews of identified contaminated sites, a further indication of loosening environmental regulation.^[12]

Executive Order “Promoting the Export of the American AI Technology Stack” (July 23, 2025) grants 90 days to the Secretary of Commerce, in consultation with the Secretary of State and the Director of OSTP, to establish and implement the American AI Exports program.^[13]

What Should You Be Monitoring?

Looking Ahead:  The Plan can at times appear daunting in its complexity.  The first Pillar alone includes more than fifty recommended policy actions aimed at accelerating the adoption of AI tools by American industry.  It will take time for the agencies tasked with carrying out these recommendations to provide further guidance, and we will continue to monitor these developments.  Nevertheless, there are several key considerations that you should have on your radar.

Another Lens for the Bias Analysis:  The AI Action Plan recommends federal procurement guidelines to grant contracts only to frontier LLM developers “who ensure that their systems are objective and free from top-down ideological bias.”  This approach may increase focus on auditing bias in controls implemented by developers within the AI system (e.g., guardrails added in system prompts), as opposed to “bottom-up” bias that arises in LLM responses because of fundamental issues with the underlying data used to train the models.  Organizations should continue to monitor any additional regulation of bias introduced by system controls and consider adding further documentation of controls to address bias, including reviews of whether these controls introduce their own levels of bias.  This is especially crucial if you sell AI to the government – prepare for potential neutrality certifications or testing as part of federal procurement.

Ongoing Analysis of Complex Compliance Landscape:  Although the Action Plan aims to loosen regulations within the federal government, it did not recommend the creation of new comprehensive AI regulation, leaving organizations in the position that they must continue to assess a fragmented compliance landscape.  Such fragmentation may result in additional regulatory obligations in other ways.  For example, the federal laissez-faire approach to AI could set up a clash with stricter state AI regulations, requiring companies with federal and state contracts to create a patchwork of systems and internal policies.  Organizations will still need to comply with existing and new state and international laws on artificial intelligence.

Monitor Updates to NIST Guidance:  NIST guidance documentation has also served as an important resource for private organizations seeking to develop AI governance programs.  While the AI Action Plan specifically directs NIST to “eliminate references to misinformation, Diversity, Equity, and Inclusion, and climate change,” organizations should continue to monitor whether NIST revises its guidance in other areas as well, including efficient and effective ways to manage risk.^[14]

Monitor Advancements in AI interpretability, Control, Robustness, and Transparency:  The AI Action Plan includes policy recommendations for programs, funding, and hackathons to advance capabilities in these areas.  Although they will take time to develop, effective AI governance programs should periodically monitor these changes for potential impacts they may have.  Such ongoing assessments of these aspects will assist organizations in assessing risk.

Monitor Creation of AI Evaluation Ecosystems and Procurement Toolboxes:  The AI Action Plan recommends rigorous evaluation in regulated industries and calls for the creation of procurement toolboxes that can be used across agencies.^[15]  Organizations should monitor the success of these programs in developing uniform and reliable approaches to assess risk associated with internal development and controls, as well as risks associated with the implementation of third-party tools.

Monitor Impact to Rules of Evidence and Procedure:  The AI Action Plan provides recommendations regarding formal guidelines and benchmarks to help identify deepfakes in forensic evidence, as well as a recommendation to adjudicatory agencies to explore updates to evidentiary rules to address deepfakes.^[16]  These recommendations serve as a reminder that courts at all levels are analyzing similar issues and that changes should be monitored.

Conclusion

The AI Action Plan introduces a new era of U.S. AI policy based on the premise that deregulation will spur innovations.  It envisions an ensuing rapid expansion of necessary infrastructure like chip manufacturing and data centers built by a trained American workforce and fueled by expanded energy capacity.  The Plan also marks an ideological shift and calls for an “ideologically neutral” approach to AI.  The Plan is also ambitious in exporting both the American AI technology stack and American standards to establish the U.S. as a global leader in partnership with allies.  Over the coming months, Redgrave will continue to monitor the implementation of these policy recommendations and provide updates on further developments.

For additional information on this topic, please contact Ray Mangum, Michael Kearney, or Jason Dollar.

The views expressed in this article are those of the authors and do not necessarily represent the views of their law firm or any of its clients