Executive Power and AI Governance: Deregulation in U.S. Policy

As artificial intelligence (AI) and machine learning technologies advance at an unprecedented pace, they usher in a new era of innovation while raising critical concerns about safety, security, and governance. The federal government plays a pivotal role in shaping the trajectory of AI development, by fostering technological progress and establishing safeguards to protect users, businesses, and national security. Central to establishing these standards are executive orders, which serve as a primary mechanism for guiding federal agencies in research, regulation, and deployment, ensuring that AI aligns with national interests. Regulatory policy, as dictated by directives such as executive orders, plays a crucial role in balancing global technological leadership with the responsible oversight of AI, especially as these systems become more deeply integrated into individual, corporate, and governmental operations [1]. Establishing clear safety and security standards early in AI’s development is essential to mitigating risks and maximizing its benefits. Given the rapid evolution of AI, government policies set today will have long-lasting implications for innovation, economic competitiveness, and ethical considerations.

Both the Trump and Biden administrations have recognized the need for federal action on AI, though with differing priorities. While Trump’s first and second administration policy has emphasized maintaining U.S. dominance in AI innovation, Biden’s directives have placed a greater focus on ethical considerations and regulatory oversight. Examining these two approaches provides insight into the evolving role of government in shaping AI’s future. 

Former President Joe Biden’s 2023 Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence—recently revoked under Trump’s second administration —places a strong emphasis on mitigating the risks associated with irresponsible AI deployment while promoting ethical and responsible development [2]. With AI capabilities advancing at an unprecedented pace, the order acknowledges the dual challenge of fostering innovation while ensuring that AI technologies do not compromise national security, civil rights, or economic stability. 

The directive establishes eight guiding principles to shape the responsible use and governance of AI, addressing safeguards for AI systems, the capacity for AI to reinforce systemic inequalities, consumer protection law compliance, data security and privacy, the capacity of the Federal government to oversee AI advancements, and U.S. leadership in technological innovation [2]. The implementation of the executive order by government departments include:

 

• A Memorandum on Advancing the United States’ Leadership in Artificial Intelligence; Harnessing Artificial Intelligence to Fulfill National Security Objectives; and Fostering the Safety, Security, and Trustworthiness of Artificial Intelligence, intended to inform future action on AI and advance the goals and objectives of the executive order [3]. 
• Implementation of the Defense Protection Act, which empowers the Department of Commerce to solicit reports on safety testing and other important information from companies developing potential dual-use foundation models, or models that have the potential for both beneficial and harmful capabilities [4]. 
• The creation of the Artificial Intelligence Safety Institute Consortium (AISIC) by the National Institute of Standards and Technology (NIST) to bring together key AI stakeholders in developing safety standards and guidelines. The consortium plays a crucial role in facilitating knowledge sharing and collaboration across sectors to support federal efforts in ensuring “safe and trustworthy artificial intelligence.” Its members are responsible for creating evaluation benchmarks, testing guidelines, and secure development frameworks to enhance AI reliability and security [5].

The executive order and subsequent policies aim to strike a careful balance: implementing safeguards that ensure AI systems are developed safely while maintaining momentum in the global race for AI innovation. Safe development and technological progress are not mutually exclusive—together, they enable the responsible adoption of AI at scale, ideally boosting efficiency and economic activity. This balance is crucial for sustaining U.S. leadership in technology. Government intervention is uniquely important in maintaining this equilibrium, given its ability to enforce standards, coordinate among companies, the public, and other stakeholders, and leverage access to cutting-edge information and infrastructure. Initiatives such as AISIC exemplify this role, offering both necessary oversight and a forum for collaboration among industry leaders and key actors.

In contrast, during both of his presidential terms, President Trump pursued a policy approach that leaned more heavily toward accelerating technological development, often by minimizing regulatory safeguards. This strategy is based on the premise that reducing oversight can accelerate innovation. While this may yield short-term gains, it introduces significant risks, particularly the development of AI systems that lack adequate safety checks. Over time, the absence of rigorous standards may lead to technical failures or system misalignments that undermine national progress rather than advance it.

Looking retrospectively, in his first term in office, President Donald Trump aimed to strengthen the United States’ role in global AI leadership, issuing the executive order, Maintaining American Leadership in Artificial Intelligence. This directive, recognizing the transformative potential of AI, prioritizes economic competitiveness and national security by expanding the Federal Government’s role in AI research, development, and deployment. Central to this initiative is the strategic allocation of federal resources to accelerate AI innovation, ensuring that the U.S. remains at the forefront of technological advancements [6]. This executive order set forth the creation and adoption of guidelines for the use of AI in federal intelligence and defense initiatives, such as:

 

• Ethical Principles for Artificial Intelligence by the Department of Defense [7]
• Artificial Intelligence Ethics Framework for the Intelligence Community by the Director of National Intelligence [8]
• Principles of Artificial Intelligence Ethics for the Intelligence Community by the Director of National Intelligence [9]

 

In addition to these defense and intelligence directives, Section 2 of the executive order emphasized cross-sector collaboration in AI research and development, encouraging partnerships among government, academia, and industry. It aimed to reduce regulatory barriers, streamline AI integration, invest in workforce development through education and training, and establish rigorous standards to maintain U.S. leadership while addressing technical security and building public trust.Furthermore,Trump issued another executive order on AI in December 2020, Promoting the Use of Trustworthy Artificial Intelligence in the Federal Government, providing additional principles for government implementation and utilization of AI, outside of national defense and security [10]. 

On his return to the Oval Office, President Trump issued a new executive order, Removing Barriers to American Leadership in Artificial Intelligence, reinforcing the goals of his first administration by prioritizing U.S. dominance in innovation and technology. However, this order shifts focus from government employment of AI towards emphasizing the removal of policy barriers to AI development and deployment. Notably, it revokes the Biden Administration’s 2023 Executive Order on AI safety, signaling a significant shift in federal AI policy [11]. While the full impact of this revocation remains to be seen, it raises critical questions about the trade-offs between accelerating AI advancement and maintaining regulatory safeguards for safe and ethical AI systems. This policy direction underscores an ongoing tension in AI governance—whether prioritizing global leadership in AI should come at the expense of oversight and risk mitigation. 

One of the most pressing challenges within this governance debate is ensuring that AI systems do not deviate from intended human objectives. A key component of AI safety that this policy may influence is alignment—the process of ensuring that AI systems operate in accordance with human values and intended goals. AI alignment is essential for developing technology that is both useful and safe, requiring continuous evaluation and refinement throughout the system development lifecycle. This involves extensive research, rigorous training, and systematic testing, making government policy a critical factor in shaping AI’s direction and safety standards. However, a major challenge in AI alignment is the diversity of human values and goals [12]. What is considered ethical or beneficial can vary widely among individuals, organizations, and governments. Policy decisions, shaped by the values of the administration or governing body, play a significant role in determining which alignment standards are prioritized and enforced.

The case of United States v. Meta Platforms, Inc. f/k/a Facebook, Inc. illustrates the real-world consequences of misaligned AI. The lawsuit alleged that Meta’s ad-targeting algorithm used Fair Housing Act (FHA)-protected characteristics, such as race, sex, religion, and disability, to determine housing ad eligibility—potentially reinforcing discrimination. Under the settlement, Meta must (1) discontinue the use of this discriminatory algorithm, (2) develop a new system to address racial, ethnic, and gender disparities introduced by its AI, and (3) eliminate FHA-protected characteristic-based targeting options for housing advertisers. Additionally, Meta must pay a $115,054 civil penalty and undergo ongoing compliance checks [13]. This case highlights the potential for AI systems to cause harm when alignment processes fail, underscoring the need for robust oversight, clear regulatory standards, and continuous monitoring to prevent AI from perpetuating bias and discrimination. Additionally, this case marks the first time the United States has challenged algorithmic bias under the FHA—an action that reflects the Biden Administration’s heightened regulatory focus on AI and the push for accountability and safety outlined in its executive order [14].

While this settlement takes significant steps to prevent Meta from employing discriminatory AI systems, the case underscores AI’s potential to perpetuate bias, particularly in the absence of proactive federal regulations. The revocation of Biden’s executive order—without a robust policy framework to replace it, especially at the training stage of AI development—heightens the risk of deploying misaligned systems. These systems can perpetuate harmful biases, particularly against Black individuals, who have already been affected by biased hiring algorithms, discriminatory housing advertisements, and unfair law enforcement profiling driven by flawed AI models. 

A recent complaint, filed by the Equal Rights Center against Meta Platforms, Inc. on February 11, 2025, serves as a case for the issue of racial algorithmic bias. The complaint alleges that Meta’s advertising algorithm disproportionately steers for-profit college ads to Black users while showing public college ads to white users [15]. The complaint’s implications for bias suggest that Black users are effectively denied access to diverse higher education opportunities, reinforcing digital redlining—the discriminatory practice of restricting marginalized communities’ access to resources through digital content, algorithms, and online platforms [16].

Although regulatory shifts following Trump’s most recent executive order are already underway, such as the Department of Energy’s initiative to use federal land for data centers supporting the growing demand for energy within the AI infrastructure, we are still awaiting the full development of regulatory policy under these new federal directives. Already, many state governments are creating regulatory legislation and policy for AI, such as  Colorado Senate Bill 24-205, which establishes consumer protections against “high-risk systems” prone to algorithmic discrimination [17], and Georgia House Bill 147, which mandates the Georgia Technology Authority to develop AI governance policies for state agencies [18]. While these state-led initiatives are crucial, the absence of unified federal standards risks a fragmented regulatory landscape. This contrast is especially evident when comparing the Biden and Trump administrations: where Biden’s approach underscores coordinated oversight and regulatory safeguards, Trump’s prioritization of rapid innovation often came at the expense of meaningful guardrails. The implications of such deregulation—potential system failures, public mistrust, and uneven industry practices—highlight the need not only for sustained centralized governance, but for federal policy that understands the role of regulation in fostering both safety and long-term innovation. Further, efforts like those of the DOE raise concerns about the environmental impacts of Trump’s deregulatory approach to AI policy, particularly in terms of increased energy consumption and emissions [19]. A comprehensive federal policy could provide clearer guidelines, ensuring the ethical development and deployment of AI systems that do not reinforce systemic discrimination.

[1] U.S. Department of Justice, Bureau of Justice Assistance. n.d. Executive Orders. Privacy & Civil Liberties.

[2] The White House. 2023. Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence. October 30, 2023.

[3] The White House. 2024. Memorandum on Advancing the United States Leadership in Artificial Intelligence: Harnessing Artificial Intelligence to Fulfill National Security Objectives and Fostering the Safety, Security, and Trustworthiness of AI. October 24, 2024.

[4] Federal Register. 2024. “Dual-Use Foundation Artificial Intelligence Models with Widely Available Model Weights.” February 26, 2024.

[5] National Institute of Standards and Technology. n.d. Artificial Intelligence Safety Institute Consortium (AISIC). Accessed April 15, 2025.

[6] Federal Register. 2019. “Maintaining American Leadership in Artificial Intelligence.” February 14, 2019.

[7] U.S. Department of Defense. 2020. DoD Adopts Ethical Principles for Artificial Intelligence. February 24, 2020.

[8] Office of the Director of National Intelligence. 2020. Artificial Intelligence Ethics Framework for the Intelligence Community.

[9] Office of the Director of National Intelligence. 2020. Principles of Artificial Intelligence Ethics for the Intelligence Community.

[10] Federal Register. 2020. “Promoting the Use of Trustworthy Artificial Intelligence in the Federal Government.” December 8, 2020.

[11] Federal Register. 2025. “Removing Barriers to American Leadership in Artificial Intelligence.” January 31, 2025.

[12] World Economic Forum. 2024. AI Value Alignment: How We Can Align Artificial Intelligence with Human Values. October 2024.

[13] U.S. Department of Justice. n.d. United States v. Meta Platforms, Inc. f/k/a Facebook, Inc. (S.D.N.Y.). Accessed April 15, 2025.

[14] Clearinghouse. n.d. Case: United States v. Meta Platforms, Inc. Accessed April 15, 2025.

[15] Equal Rights Center. 2025. Complaint. February 11, 2025.

[16] National Center for Biotechnology Information. 2022. Article PMC9339607.

[17] Colorado General Assembly. 2024. Senate Bill 24-205.

[18] Groves, Caleb. 2025. “Georgia Legislators Push for More AI Guardrails.” Government Technology, March 27, 2025.

[19] American Council for an Energy-Efficient Economy. 2025. Future-Proof AI Data Centers, Grid Reliability, and Affordable Energy: Recommendations for States. April 7, 2025.