The Policy Tug-of-War: America s Pivotal Shift in AI Regulation

I. Lead Section: The Regulatory Backlash

The governance of Artificial Intelligence in the United States is currently defined by volatility, following a rapid and decisive reversal of federal policy that has reshaped the national strategy. After a period of incremental consensus-building that culminated in a comprehensive safety mandate, the federal government executed a profound pivot away from mandatory risk management toward accelerating technological dominance.

The core of this shift lies in the immediate rescission of the Biden-era Executive Order 14110 (EO 14110). This landmark order had focused on promoting the "Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence," outlining a national approach to governing AI and specifically targeting the prevention of AI-enabled threats to civil liberties and national security. Within days of the subsequent administration taking office, EO 14110 was revoked.  

The successor policy, framed by Executive Order 14179 and the subsequent America's AI Action Plan, explicitly shifts the federal focus onto "Removing Barriers to American Leadership" and achieving global AI dominance. The stated objective is to lessen potential bureaucratic burdens and restrictions that, according to the administration, have "hindered timely uptake of AI across federal agencies". This reorientation signals a clear prioritization: speed, economic competitiveness, and technological acceleration have been placed ahead of centralized, mandatory risk management requirements.  

The immediate and comprehensive nature of the EO 14110 rescission created a significant regulatory shockwave and a compliance vacuum across the executive branch. The prior order required extensive implementation across agencies regarding high-risk AI accountability and reporting. The new executive directive requires agencies to "suspend, revise, or rescind" these prior actions. Furthermore, in cases where finalization of those changes is not immediate, agency heads are instructed to promptly provide all authorized exemptions under existing orders, rules, or policies. This directive demonstrates that the immediate policy priority is the de-escalation of compliance demands, a move which, while intended to spur innovation, potentially increases the near-term risk exposure of government systems and undermines accountability mechanisms that were already in progress.  

II. Background: A Foundation of Ambition and Caution

While executive orders can shift dramatically with political transitions, the U.S. AI governance landscape is anchored by foundational statutory law and enduring technical standards that offer stability.

The Statutory Bedrock: The National AI Initiative Act (NAIIA) The true legislative foundation for federal AI policy is the National Artificial Intelligence Initiative Act of 2020 (NAIIA). Enacted as part of the William M. (Mac) Thornberry National Defense Authorization Act for Fiscal Year 2021, the NAIIA outlined a comprehensive plan for a National AI Initiative, encouraging cross-agency collaboration and mandating reports for accountability.  

The NAIIA acknowledged fundamental structural challenges that continue to define the AI landscape. Federal guidance identified a lack of clear government understanding of AI capabilities and its potential effects on social and economic sectors, including ethical concerns, national security, and workforce impacts. Furthermore, the Act noted that researchers outside of Big Tech—in academia, federal laboratories, and many private sectors—have limited access to high-quality datasets, sufficient computing resources, or real-world testing environments necessary to design and deploy safe and trustworthy AI systems. This foundational legislative mandate for R&D, standardization, and addressing resource disparity remains stable, transcending the recent executive policy shifts.  

The Technical Baseline: NIST and the Enduring Framework for Trust The volatility inherent in executive policy is significantly counterbalanced by the enduring, consensus-driven guidance provided by the National Institute of Standards and Technology (NIST). The AI Risk Management Framework (AI RMF), issued by NIST, is a key component of U.S. governance.  

The AI RMF is a guidance document, not a federal regulation, designed to improve the robustness and reliability of artificial intelligence systems by providing a systematic approach to managing risks. It is built upon four core, iterative functions that organizations are expected to implement throughout an AI system's lifecycle :  

Govern: Emphasizes cultivating a risk-aware organizational culture, starting with leadership commitment and the establishment of clear governance structures.  

Map: Focuses on contextualizing AI systems within their operational environment, identifying potential impacts across technical, social, and ethical dimensions.  

Measure: Promotes detailed risk assessment using quantitative and qualitative approaches to understand the likelihood and potential consequences of risks.  

Manage: Guides organizations in prioritizing and addressing identified risks through a combination of technical controls and procedural safeguards.  

The framework’s distinctive socio-technical approach recognizes that AI risks extend beyond technical considerations, encompassing complex social, legal, and ethical implications. This approach requires organizations to consider a broad range of stakeholders and potential impacts when developing and deploying systems, ensuring they align with values like validity, reliability, safety, security, and resilience.  

The NIST AI RMF functions as the "shadow regulation" or operational floor for responsible AI. Given the U.S. resistance to adopting rigid, centralized AI laws (in contrast to other global jurisdictions), the NIST RMF becomes the de facto standard for organizations seeking to demonstrate due diligence and mitigate liability. This stability is crucial for industry, as technical standards and recognized methodologies for risk management often outlast political mandates, making the NIST guidance a consistent tool for aligning AI systems with organizational values and societal norms.  

III. Analysis: Policy Velocity and the Dual Mandate

The shift in U.S. federal AI strategy highlights the deeply entrenched debate over the necessary equilibrium between accelerating innovation and preemptively managing systemic risks.

The Pivot to Unfettered Competitiveness The current executive direction, articulated in the America's AI Action Plan, explicitly pursues deregulation to facilitate rapid technological advancement, aiming for global AI dominance. The policy rescinded the previous executive order, citing that it imposed "burdensome regulations that hindered innovation and undermined America's global competitiveness in AI".  

The objective is now defined by promoting competition and accelerating breakthroughs in key sectors such as medicine and manufacturing, improving the standard of living for Americans. Furthermore, the administration emphasizes that AI systems must be "free from ideological bias" and designed to pursue "objective truth," positioning trustworthiness within the context of avoiding political or social engineering agendas. Implementation requires agencies to review and, where appropriate, "suspend, revise, or rescind" any policies stemming from the revoked EO 14110 that are deemed inconsistent with the new pro-innovation mandate. The Office of Management and Budget (OMB) supports this direction, actively working to lessen bureaucratic restrictions that could impede the timely uptake of AI across federal agencies.  

The Critical Tension: Safety versus Speed This rapid policy pivot encapsulates the core governance conflict: whether regulation is a necessary prerequisite for public trust or an undue brake on technological progress.

The tension is amplified by the rapid acceleration of generative AI, which has introduced novel risks at scale, including the heightened potential for fraud, cybercrime, deepfakes, and manipulation of public opinion. This technology shock demands policy responsiveness, yet the U.S. response has been to decelerate mandatory risk management. Policy experts argue that embedding transparency, accountability, and fairness into AI systems from the outset is necessary to prevent harm—particularly to marginalized communities—and is not inherently contradictory to maintaining U.S. leadership. Effective policy relies less on technical descriptions of the world and more on terms that move stakeholders to constructive action, often centered on language of safety and responsibility.  

The Competition Paradox: Big Tech's Regulatory Moat A complex consequence of varying regulatory regimes is the impact on market concentration. Ironically, measures intended to simplify the landscape for all companies, including startups, may end up accelerating market consolidation in favor of incumbent giants.

Rigorous regulatory compliance, such as extensive reporting and vetting requirements, typically imposes substantial costs. These costs are easily absorbed by Big Tech firms (hyperscalers), but they tend to "squeeze out smaller start-ups" by diverting scarce resources away from research and development . These compliance barriers deter new entrants, leading to a narrower diversity of AI developers and potentially limiting breakthroughs. The removal of mandatory safety regulations does not automatically guarantee a more competitive marketplace; in fact, it may exacerbate the issue.  

Antitrust concerns regarding AI-driven market dominance have been raised globally by watchdogs, including the U.S. Federal Trade Commission . These bodies recognize that the concentration of data, compute resources (controlled by cloud vendors like AWS and Azure), and ownership of closed AI models by a few large players risks creating new monopolies . Startups often rely heavily on these hyperscalers for the necessary layers of software to scale quickly. This dependency creates architectural lock-in, where today's partner can quickly become tomorrow's competitor, reinforcing a centralized control over the fundamental infrastructure of the AI ecosystem . If the primary barrier to entry is capital cost and resource access—which is controlled by the hyperscalers—rather than regulatory compliance, removing safety regulation does little to solve the underlying antitrust problem. If a startup, operating under relaxed safety standards, later causes a significant public harm event (e.g., through widespread bias or misuse), public backlash will inevitably lead to retroactive, punitive regulation that only the largest firms have the balance sheets to survive, ultimately strengthening their regulatory moat.

The table below visually captures the stark and immediate philosophical shift in the executive branch’s approach to AI governance, illustrating the policy tug-of-war.

Core Policy Comparison: US Federal AI Strategy Shift (2023 vs. 2025) Policy Area Biden Administration (EO 14110) Current Administration (EO 14179 & Action Plan) Primary Focus Safe, Secure, and Trustworthy Development; Civil Liberties

Removing Barriers to Leadership; Innovation and Competitiveness

Regulatory Stance Mandates on high-risk AI, agency accountability, prevention of harm Expedited adoption, review/rescission of burdensome regulations

Guiding Principle Risk mitigation and public protection

Global dominance and economic growth, freedom from "ideological bias"

Immediate Action Mandatory risk reporting for critical AI models Promptly provide "all available exemptions" to existing compliance

IV. Sectoral Impacts: Where Regulation Endures

In the absence of a comprehensive, unified AI law, the U.S. governance model relies heavily on the decentralized resilience of specialized federal agencies enforcing existing mandates. This sectoral approach ensures that risks in high-stakes areas like finance, consumer safety, and healthcare are continually addressed, regardless of broader policy volatility.

Consumer Protection and Fraud: The FTC's Active Stance The Federal Trade Commission (FTC) remains a highly active and effective regulator in the AI space, demonstrating that existing laws are robust enough to combat AI-related harm in the marketplace . The FTC uses its authority to file suits against AI-enabled fraud and deceptive practices. For example, the commission has targeted schemes that falsely guaranteed consumers income through "AI-powered software" and fraudulent online storefronts .

These enforcement actions prove that developers and operators of AI systems are not exempt from accountability under traditional consumer protection statutes. They send a clear message that claims of "proprietary software" or "cutting edge AI-powered tools" must align with factual performance and legitimate business opportunities. Operators of fraudulent schemes have faced severe penalties, including being permanently banned from the business sector and required to turn over funds for consumer redress . The FTC's actions focus on market honesty and prevent the technology from being weaponized against individuals seeking financial gain, anchoring accountability in consumer protection law.

High-Stakes Health Applications: The FDA’s Lifecycle Management For critical fields like healthcare, the Food and Drug Administration (FDA) is advancing detailed guidance for AI-enabled device software functions. The regulatory challenge in this sector is unique because many AI models used in medical devices are designed to adapt and evolve post-deployment.  

The FDA guidance emphasizes a comprehensive approach to risk management throughout the device’s Total Product Life Cycle (TPLC), covering recommendations for design, development, implementation, and post-market oversight. This ensures that the essential framework for safety, predictability, and trustworthiness is established, regardless of the broader policy signals. For regulated products, the FDA ensures that algorithmic changes—even those implemented after initial marketing—are managed under existing premarket and post-market authority, establishing a necessary regulatory floor for patient safety.  

Government Operations: OMB Directives for Federal Agencies The Office of Management and Budget (OMB) serves a critical function in standardizing AI procurement and use within the federal government itself. The OMB memo M-24-10 aims to maximize public benefit from timely AI adoption across federal agencies while also ensuring that all AI used or acquired adheres to necessary safety and security standards .

This directive acknowledges that the federal government must leverage AI to improve public services. However, it contains a significant limitation: the guidance does not cover AI being used as a component of a National Security System . This carve-out means that defense and intelligence applications operate under separate, likely less transparent, governance structures, reflecting the complexity of managing AI across competing federal mandates.

This decentralized model of enforcement provides the U.S. with a degree of regulatory flexibility. By avoiding rigid, sector-agnostic legal definitions, the approach is less likely to become outdated and restrictive as AI evolves rapidly, a point of consensus noted in international policy discussions. This sectoral enforcement ensures that while the executive branch can signal policy preferences, core public protections related to money, health, and national security remain largely intact, enforced by the domain experts.  

V. Key Impacts: Stakeholder Consequences

The current policy environment creates distinct consequences for policymakers, industry stakeholders, and the public, highlighting systemic risk factors tied to deregulation and market dynamics.

For Policymakers and Congress Policymakers face the immediate and pressing challenge of maintaining public trust when the governmental signal suggests that accelerated adoption supersedes mandatory safety vetting . The policy language deployed to justify deregulation, which shifts the debate away from deep social concerns and toward narratives of "existential risk" or "ideological bias," serves to simplify complex risks but often obscures crucial issues related to the sociology of AI design and deployment decisions.  

Without clear statutory authority, the AI landscape will continue to be governed by the political pendulum swings of executive orders, creating unsustainable long-term uncertainty for industry and consumers alike. Congress has a vital role to play in advancing policies that can simultaneously cement U.S. leadership while protecting the public interest. Failure to achieve legislative consensus means that safety frameworks will remain vulnerable to being suspended or rescinded by subsequent administrations, undermining any long-term stability needed for responsible investment.  

For Tech Companies (Big Tech vs. Startups) The regulatory uncertainty disproportionately benefits hyperscalers (Big Tech). These companies possess the financial and engineering resources necessary to navigate shifting regulatory standards and, crucially, they control the vast compute infrastructure that forms the bottleneck for frontier model development . Due to their dominance over data and compute, Big Tech is less sensitive to compliance costs that might cripple smaller rivals .

Small AI innovators face a structural dilemma. While they are now freed from immediate federal regulatory burdens, they remain fundamentally dependent on Big Tech’s infrastructure . High compliance burdens, if they are eventually re-imposed, or market lock-in by hyperscalers, make long-term, independent growth challenging. Furthermore, antitrust bodies have flagged the risk of algorithmic collusion and gatekeeping facilitated by centralized control over AI models, suggesting that a competitive marketplace is threatened regardless of the regulatory burden .  

This dynamic implies that the U.S. strategy, while aiming for competitive freedom, risks accelerating market consolidation in favor of incumbent actors. Since Big Tech can absorb future regulatory costs and dictates access to the essential means of production, deregulation primarily shifts responsibility rather than creating genuine market parity.

For Society and the Public When the emphasis of federal policy shifts away from mandatory accountability and fairness, the potential for algorithmic bias, discriminatory outcomes, and resulting harm to marginalized communities increases . The socio-technical risks identified by the NIST RMF—which include privacy breaches, bias, and security threats—are inherently difficult to manage consistently without rigorous, enforceable oversight mechanisms.  

Furthermore, the integrity of the information ecosystem faces a severe and immediate threat. The proliferation of deepfakes and mass disinformation, heightened by the capabilities of generative AI , requires constant vigilance and effective policy response. The recent decision to move away from centralized risk management transfers the complex responsibility for addressing these systemic societal threats almost entirely to the voluntary actions and internal policies of the major technology platforms.  

The U.S. strategy risks implicitly outsourcing the crucial task of trust-building to Big Tech. By retracting mandatory safety protocols that require transparency and testing, the government is effectively transferring the burden of ensuring safety and trustworthiness onto the large companies that develop and deploy the most advanced models. This reliance is problematic due to the inherent conflict of interest between maximizing market speed and ensuring comprehensive risk mitigation. If this outsourced trust framework fails, resulting in a major social harm or system collapse, the ensuing public trust deficit will necessitate a severe legislative overcorrection, potentially leading to overly rigid laws that stifle all but the largest, most entrenched players.

VI. Conclusion: Navigating the Trade-off

The current moment in U.S. AI governance represents a sharp confrontation between two competing national imperatives: the drive for technological supremacy and the need to guarantee public safety and trust. The recent executive policy reversal, prioritizing speed and economic competitiveness, clearly articulates the administration's preference for removing perceived barriers to innovation in the pursuit of global dominance.

Yet, the comprehensive analysis reveals that the strength of American AI governance does not rest solely on the direction set by executive orders. Stability and trustworthiness are maintained by the decentralized resilience of specialized sectoral agencies—such as the FTC in consumer protection and the FDA in healthcare—and the consistent technical stability offered by the NIST AI RMF. These components provide a necessary, enduring floor for responsible AI development, ensuring that high-stakes applications remain subject to rigorous, domain-specific scrutiny.

Ultimately, regulation should not be conceptualized as an inevitable "brake" on progress, but rather as the foundation for lasting innovation. Clear, predictable rules enable organizations to manage risk and build systems that consumers trust, encouraging wider adoption. Conversely, policy volatility, compounded by the competitive paradox wherein light-touch regulation accelerates Big Tech dominance, creates an environment of systemic fragility. Until Congress achieves a clear, consistent, and durable statutory consensus that explicitly embeds transparency, accountability, and fairness into the development lifecycle, the greatest long-term threat to American AI leadership will not be foreign competition, but the internal collapse of public confidence. The AI race, in the current environment, risks being defined by short-term policy sprints and reactive damage control, rather than sustained, responsible growth.