AI Governance at a Crossroads: America’s AI Action Plan and its Impact on Businesses

corporate buildings in clouds

AI Governance at a Crossroads: America’s AI Action Plan and its Impact on Businesses 

This post is the first in a series of commentaries on recent developments in the sphere of AI governance and innovation. In this post, we examine the recent America’s AI Action Plan and its shifting priority towards a philosophy of AI deregulation. The Plan advocates for innovation over caution, while this uncertain and ever-changing legislative landscape will demand a further shift towards the private sector to manage AI risks related to ethics, responsibility, and governance in the absence of legal mandates.

Regulatory Landscape Prior to the AI Action Plan

In the United States, the first significant federal action on AI governance came during President Trump’s first administration, with the 2019 Executive Order on Maintaining American Leadership in Artificial Intelligence. The order focused on promoting U.S. competitiveness by directing federal agencies to prioritize AI R&D, workforce training, and international cooperation. It framed AI largely as an engine for innovation and economic growth but offered little in the way of oversight or guardrails.

Building on this foundation, President Biden’s 2023 Executive Order on Safe, Secure, and Trustworthy AI shifted the tone considerably. This order signaled that AI was not just an economic driver, but a matter of civil rights and national security, in need of guidelines. The order emphasized the importance of ensuring AI systems are safe and equitable while encouraging innovation and safeguarding America’s competitive edge. The order directed agencies to adopt the NIST AI Risk Management framework as the standard for AI oversight, and agencies began developing principles and best practices for use of AI within their domains, covering areas such as data privacy, workplace monitoring, and protection against algorithmic discrimination. 

At the same time, the absence of comprehensive federal legislation opened the door to a growing wave of state-level activity, with now over 480 enacted bills referencing “artificial intelligence”. States such as New York, Colorado, and California introduced their own AI-related laws and proposals, ranging from hiring bias audits to algorithmic discrimination bills. These efforts, combined with a lack of clear federal law, created a fragmented and inconsistent regulatory landscape making it challenging for companies operating across jurisdictions to anticipate potential liabilities and comply with competing standards.

When Trump came into office for his second term, President Biden’s Executive Order was lifted, with the intention of making unhindered innovation a priority. Subsequent efforts to streamline oversight, including a proposed moratorium on state-level AI regulation, did not gain traction. As a result, by 2025 the U.S. regulatory environment remained characterized by gaps, overlaps, and uncertainties. This unsettled foundation set the stage for the development of a more comprehensive national framework through the AI Action Plan.

The AI Action Plan – Supporting Innovation through Deregulation

On July 23, 2025, the White House released America’s AI Action Plan, outlining the Trump administration’s federal AI strategy in its effort to win the AI race against its closest competitor, China. Alongside this Plan, the administration also released a series of three executive orders that align with this national promotion of AI innovation:

  1. Preventing Woke AI in the Federal Government,
  2. Accelerating Federal Permitting of Data Center Infrastructure, and
  3. Promoting the Export of the American AI Technology Stack.

These executive orders reflect a broader shift in U.S. policy, where the central driver of AI governance focuses less on ethical safeguards and more on maintaining a competitive edge on the world stage. The Plan frames AI innovation as a domain of economic dominance and national security in which the AI race is presented not simply as a matter of technological progress, but as essential to protecting U.S. influence in international markets and strategic alliances. 

The Plan is structured around three main pillars: 

  1. Accelerating Innovation,
  2. Building American AI Infrastructure, and
  3. Leading in International Diplomacy and Security. 

At its core, this national vision adopts a pro-innovation approach to AI regulation, aiming to ensure that the United States remains the commanding force in global technology norms and standards. This approach involves prioritizing technological growth and development over precautionary oversight, while also sending a strong message to both allies and competitors alike about American intentions. More broadly, this shift in regulatory strategy underscores the volatility of the American regulatory landscape as shaped by changes in administration.

The Plan advocates for reduced federal regulation of AI systems and use cases, which, as we argue further, would place greater responsibility on corporate boards, officers, and senior management teams, to self-manage and mitigate AI risks within their own organizations. This deregulatory stance is clear from the outset of the document, as the first section of the first pillar is titled “Remove Red Tape and Onerous Regulation.” Its goal is to allow the private AI sector to flourish free of what it describes as the “bureaucratic red tape” imposed under the Biden administration. Central to this vision is rolling back existing policies and frameworks such as Biden’s Executive Order 14110 described previously. The Plan further recommends withholding federal funding from states with “burdensome” AI regulatory frameworks, while still allowing states to pass laws so long as they do not obstruct technological progress. This approach sets the stage for a fragmented and largely deregulated AI landscape across the country.

The conclusion of this section lists recommended policy actions to support deregulation, beginning with an investigation into current policies that impede AI innovation and adoption led by the Office of Science and Technology Policy. Additional recommendations task the Office of Management and Budget and the Federal Communications Commission with revising regulations that conflict with these priorities and aligning federal funding with states’ regulatory climates. Although many of these recommendations lack a clear timeline or funding plan, raising questions about the pace and extent of implementation, it is apparent that greater changes in the U.S. regulatory landscape both at the federal and state level should be expected soon. At the state level, legislative actions have the potential to be swayed by bipartisan alignment with either this new innovation-driven approach from the Trump administration or greater attention to values based principles set forth under the Biden administration.
Subsequent sections of the Plan re-emphasize the administration’s pro-innovation stance while also maintaining the role of the federal government to manage risks involving technological standards and model evaluations across a variety of government departments. According to a recent Stanford HAI report, this strategy can be thought of as a shift from prescriptive regulation towards evidence-based policymaking. While there are calls for greater security and risk management frameworks, the Plan notably lacks federal commitment to mitigating ethical concerns around bias, transparency, and data privacy that characterized the Biden Administration’s approach. As in the opening section, many of the following proposed actions also lack specificity and concrete pathways for implementation.

Notably, just three days after the Trump administration’s Plan was published, the Chinese government released a responding Global AI Governance Action Plan, emphasizing fairness, cooperation, national sovereignty, and the service of humanity at the forefront of its AI goals. This back-and-forth of policy initiatives heightens the AI race between the United States and China that characterizes much of the Trump Administration’s regulatory frameworks. As explored in one of our previous posts, these tensions have been building gradually as policymakers across the globe are trying to find the right way to balance innovation with the need to protect human values and guard against risks.

Growing Pressure on Corporate Leaders to Govern AI

Within our research, we have established the immature and volatile AI regulatory landscape as a major motivator for corporate action. As the federal government shifts from prescriptive regulation to “light-touch” governance and state-level legislators are incentivized to move in a similar direction, the responsibility increasingly falls to the private sector to establish and maintain its own standards of governance as they are not able to rely on governments to provide effective risk-management frameworks. This raises a critical question about what motivates companies to spend the time and resources on ethical AI oversight systems without external compliance motivators. 

National deregulation does not eliminate the risks that AI poses to businesses, including reputational, operational, financial, strategic, and data security risks that remain significant and that require ethical AI frameworks to mitigate them. As we’ve explored in a previous blogpost, there are numerous examples of misaligned AI systems and governance structures that have led to a variety of negative impacts across companies. Therefore, we believe that the implementation of ethical AI  governance practices is more than just a moral obligation, but rather a necessity for companies to avoid a variety of risks, especially financial and reputational risks. Moreover, existing legal precedents, such as the Caremark progeny of cases that we discussed in a previous blogpost, maintain legal liability for boards and executives to provide sufficient oversight of risks tied to “mission critical” company operations that increasingly involve AI systems. This can be achieved by implementing necessary guardrails within organizations to ensure safe, secure, and trustworthy AI development and deployment. 

Throughout our research we’ve developed an AI governance framework – the Boundaries of Tolerance Framework – that revolves around determining the boundaries of what risks, ethical considerations, and outcomes each specific organization will tolerate in their AI development and deployment. Our research uses AI and business ethics as mediums to achieve this goal amongst an evolving national regulatory landscape. 

Ultimately, the shifting nature of U.S. national regulation places increasing pressure on corporations to self-manage their AI practices and to develop robust AI-management frameworks. Within the environment of evolving state-level legislative frameworks and federal executive branch policies it is critical for companies to remain informed and hold themselves accountable to manage not only potential legal compliance, but also to manage broader ethical, operational, and reputational risks.

***

  • Jeffrey Saviano: Business AI Ethics Initiative Leader; Edmond & Lily Safra Center for Ethics, Harvard University | jeffreysaviano@fas.harvard.edu
  • Jonathan Hack: Director of Content & Strategy; Edmond & Lily Safra Center for Ethics, Harvard University | jhack@fas.harvard.edu
  • Vija Kalnina, PhD; Business AI Ethics Initiative Lead Researcher
  • Lily Noyes; Business AI Ethics Initiative Research Assistant
  • Ryan Wettre; Business AI Ethics Initiative Research Assistant

The Business AI Ethics research team is part of the Edmond & Lily Safra Center’s ongoing effort to promote the application of ethics in practice. Their research assists business leaders in examining the promise and challenges of AI technologies through an ethical lens. Views expressed in these posts are those of the author(s) and do not imply endorsement by ELSCE.