1. Introduction
In January 2025, just hours after taking office, President Trump revoked former President Biden’s Executive Order (EO) on Artificial Intelligence (AI) by issuing a new EO titled “Removing Barriers to American Leadership in Artificial Intelligence”. [1] During his first term, Trump issued two EOs[2] that set out principles for the government’s use of AI, especially safety, security, monitoring, and respect for civil liberties.[3] The interest in civil liberties was notably absent in the 2025 Trump EO. Biden’s now-revoked EO aligned with such principles, encouraging AI innovation and security in the US by emphasizing voluntary guidelines, economic competitiveness, and industry self-regulation. Despite this continuity, Trump’s revocation is unsurprising because his original EOs covered only the federal government’s use of AI, while Biden’s EO extended to the private sector — a move likely perceived as an overreach by the Trump administration. Biden’s EO was built on existing commitments from major tech firms—including Amazon, Google, Meta, Microsoft, and OpenAI—to undergo third-party oversight,[4] but required companies to share details about their most powerful AI models with the government before releasing them to the public.
The 2025 EO directs the creation of an AI Action Plan, which a White House-led Interagency task force is currently formulating.[5] Also, President Trump established a Presidential Advisory Commission on Science and Technology on Inauguration Day, uniting leaders from government, industry, tech, and academia to ensure that the US remains a global leader in science and technology.
It remains unclear what additional federal actions the new administration will take, or how they might affect individual states and existing state law. In the absence of a clear policy from the current administration, Vice President J.D. Vance’s recent remarks at the Paris AI Summit on “excessive” regulation stifling innovation are informative.[6]This view aligns with the Trump EO’s vision to avoid “unnecessarily burdensome” obligations for AI companies.[7]
However, many US states have recently taken a proactive stance on AI regulation, making it difficult for the Trump administration to undo these efforts overnight. In 2024 alone, nearly 700 AI-related bills were introduced, targeting issues like algorithmic bias, transparency, and AI-generated content, with Colorado and California leading with specific proposals.[8] Given that some of these proposed or enacted laws do not align with the laissez-faire approach that the current administration is poised to adopt, potential conflicts between the legislative powers of the federal government and the states are likely. This is the topic of this article in which we explore two main scenarios for AI regulation in the United States. In one, states continue to enact patchwork regulations with minimal or no federal intervention. In the other, the federal government actively restricts patchwork state action to maintain its proposed laissez-faire pro-industry approach across the board. In discussing these scenarios, we highlight a critical challenge in the current administration’s proposed AI agenda. With a push for deregulation, the Federal Government ostensibly aims to limit state law-making. Yet, as existing practice shows, any effort to limit state law requires the Federal Government to accept some degree of legislative interventions or cohesive congressional consensus. Both are difficult to achieve.
2. Two scenarios after Trump’s 2025 EO
In 2019, the Trump administration’s American AI Initiative to develop technical standards[9] and secure US AI leadership was criticized for its vague goals and insufficient funding.[10] The Biden administration took a more proactive approach with the AI Bill of Rights, increased R&D funding, and the now-revoked executive order.
The current administration’s EO emphasizes the need for the country to maintain global leadership by ensuring that AI systems remain free from ideological bias or engineered social agendas. However, it presents a mix of seemingly contradictory statements. Some sections apparently strive for neutrality, stressing the need to eliminate ideological bias to foster free speech and an entrepreneurial spirit. In contrast, others take a more assertive tone, for example, the call to “sustain global AI dominance to promote human flourishing”.
More concrete insights can be found in the accompanying fact sheets.[11] There, Biden’s EO is characterized as a threat to AI innovation and criticized for imposing “unnecessary government control over the development of AI” in the private sector and establishing excessively burdensome requirements for companies. Although no information is provided about which requirements are deemed overly restrictive, it is reasonable to infer that measures like mandatory information sharing, compliance with development conditions, and rigorous testing and post-deployment performance monitoring—such as the “AI red-teaming” tests mentioned in Biden’s Order—were perceived as too demanding. This critique should be interpreted within the broader context of differing rhetorical styles: the current administration’s directives, including the 2025 EO and those from his first administration, strongly emphasize American leadership, while Biden’s approach placed significant focus on international collaboration and global partnerships.
The text of the Trump EO and accompanying fact sheets do not state whether the federal government intends to legislate. It calls on agencies and departments to “revise or rescind” any policies, regulations, or directives that no longer align with the current government’s AI leadership mandate.[12] The challenge is that, if federal legislation is introduced, it could easily conflict with states’ autonomy to regulate AI, particularly in light of the current administration’s apparent preference for deregulation. At the same time, if no federal legislation is introduced, state-level regulations may depower the new EO’s, making it ineffective. In this context, two scenarios should be considered to make some sense of the complex and evolving regulatory landscape of AI: (a) decentralized state-level regulation without federal interference and (b) federal dominance via pre-emption of state law.
2.1. Scenario 1: Decentralized state-level regulation
In the first scenario, the Trump Administration’s new EO and the ensuing future directives may leave room for states to legislate how they see fit, as the federal government’s guidance is deliberately minimal. This means that AI regulation could develop in ways similar to data protection, through a coordination of state-level efforts facilitated by the National Conference of State Legislatures (NCSL).
Ironically, the practical effect of this approach would not be much different from the previous administration’s status quo. Federal activity on AI legislation was relatively limited during the Biden administration, owing to the lack of Congressional consensus and the reluctance to regulate an evolving field. Most concrete policymaking occurred at the state level. For example, the Colorado Privacy Act (CPA) imposes obligations on businesses of all sizes—including some AI model developers outside the state—processing personal data, with provisions about profiling and heightened responsibilities for specific data uses.[13] Likewise, Connecticut tried to pass a comprehensive bill regulating private sector AI use, using similar terminology to the AI Act (e.g., referring to high-risk systems), but the legislation stalled due to tech industry opposition.[14] Finally, Tennessee enacted the “ELVIS Act,” intended to shield recording artists’ names, likenesses, and voices from AI-generated deepfakes.[15] Many more states are planning similar initiatives.[16]
State initiatives have been limited in scope, concentrating on preventing discriminatory uses of AI, safeguarding intellectual property, combating deepfakes, and mitigating additional risks posed by generative AI. However, some states pursue a more ambitious path, trying to regulate AI at the model level. For instance, California—where about 70% of the world’s leading AI companies are based[17]—attempted to mandate safety measures for developing AI systems and even establish a dedicated agency to oversee the development of any AI model “distributed” in California. Although the bill was ultimately vetoed, it highlights how regulatory efforts, due to the inherently borderless nature of digital technologies, could exert influence beyond state lines and grant broader oversight to such agencies.[18]
In this scenario, state-level regulatory autonomy allows tailored rules that address local needs and priorities. AI’s broad applicability across diverse sectors means there are many policy areas where states should legitimately enforce their priorities and values regarding its use and the ensuing liabilities.[19] For example, states may differ significantly in regulating AI in medical applications. One state might prioritize patient privacy and require strict limitations on how AI systems access and use health data, while another might emphasize innovation and allow broader use of AI in diagnostics and treatment planning to improve healthcare outcomes. These differences reflect each state’s unique preferences, highlighting the importance of preserving state flexibility alongside federal oversight to ensure that AI regulation is effective and responsive to local needs.
Given federal deadlocks on AI policy and the lack of consensus on a comprehensive or sectoral AI law, state action is usually faster and more responsive to evolving industry challenges. This flexibility ensures oversight and safety without a coherent federal strategy. Ideally, some virtuous states with forward-thinking policies and effective governance may serve as forerunners for other states, because state legislators try to coordinate and harmonize their efforts and catalysts for federal action, highlighting pressing issues and successful models to inform national policy development. This will involve identifying a state law that could serve as a minimal yet effective national standard and working to ensure its adoption across the country.
At the same time, this autonomy may lead to legislative fragmentation and potential enforcement disputes. As mentioned, many states are already taking steps (or attempting to) toward stricter oversight and explicit ethical guidelines. In contrast, others may adopt more permissive policies or choose not to act to attract AI investment. This patchwork of regulations introduces heterogeneous approaches to privacy, civil rights, risk management, and trade-offs between these norms. As a result, businesses operating across multiple states may face varying obligations, heightening compliance complexity and associated costs.
This decentralized environment could also generate novel legal tensions. For instance, Tennessee’s ‘ELVIS Act’ aimed at protecting musicians’ voices and images from unauthorized use by AI, could face First Amendment challenges, as courts may need to balance publicity rights against free speech protections. Some critics argue that the ELVIS Act’s expansive scope, which includes liability for making an individual’s voice or likeness available to the public, could create chilling effects on fair use creative endeavors—e.g., parodies and even news reporting—and place undue financial and legal burdens on smaller creators.[20] While states take different approaches to publicity rights, these variations do not always result in direct legal conflicts. However, challenges may still arise in determining which state’s law applies, depending on factors such as where the affected individual resides or where the alleged violation occurred.
Another concern is that the lack of uniform standards may impede AI deployment and economy of scale in heavily regulated industries such as healthcare, finance, and public safety. Because similar disputes could see different outcomes across states, forum shopping—plaintiffs filing suits in jurisdictions known for plaintiff-friendly or defendant-friendly rulings—may become more likely.
Furthermore, challenges could arise from the Commerce Clause if corporations argue that state-level AI regulations unduly burden interstate commerce. Courts would then have to weigh each state’s power to protect local interests against the broader need for national uniformity in a rapidly evolving, tech-driven market.
Finally, questions remain about whether states have the financial and administrative capacity to implement and enforce their AI regulations effectively.
2.2. Scenario 2: Federal Dominance
In this section, we outline the alternative scenario in which the current federal administration seeks to dominate state-level AI legislation using the tools at its disposal.
As mentioned, the new Trump EO mandates the development of a comprehensive AI Action Plan spearheaded by leading federal science and national security officials. According to the administration, this plan is designed to ensure that models remain free from ideological bias. The EO also directs the White House Office of Management and Budget (OMB) to update federal AI governance guidelines to foster greater flexibility in collaboration with the private sector, thereby eliminating unnecessary obstacles to America’s leadership in AI innovation. Given the history of bipartisan legislation proposed in Congress, the administration could push comprehensive model safety laws or sectoral laws to remit antitrust, speech harms (deepfakes in particular), and intellectual property.[21]
The Biden Administration’s cautious approach towards concrete action can be attributed to the uncertainty of AI’s impact and its evolving benefits and risks, efforts made by powerful AI lobbies towards deregulation, and the lack of consensus among lawmakers. Instead, the Trump Administration may push its current vision of AI deregulation or selective federal regulation at the state level. A patchwork of state laws not only hinders the current administration’s efforts to bolster US dominance in AI but also creates significant business uncertainty for the homegrown AI industry.
One way the federal government can achieve its vision is through “pre-emption” of state law—the displacement of state law by federal statutes in force of the Supremacy Clause of the Constitution (Article VI, Clause 2).[22] This measure can help federal lawmakers curtail state-level AI regulations and reserve key policy areas for the federal government. However, as discussed below, the doctrine is not a green light for federal exclusivity. Given the nature of AI and its risks, it will trigger complex statutory and jurisprudential challenges. At this point, because of the novelty of technology and legislation in the field, it is difficult to predict how this would play out in specific sectors. The following paragraphs provide a broad conceptual overview.
The Supremacy Clause serves as the constitutional foundation for pre-emption. Courts evaluate whether pre-emption applies on a case-by-case basis and generally operate under a presumption against pre-emption. This presumption, which is a canon of statutory interpretation[23], reflects the judiciary’s respect for the traditional police powers of the states and the principle of dual sovereignty. In areas traditionally regulated by states, such as health, safety, and welfare, courts will not infer pre-emption unless Congress’s intent to pre-empt state law is warranted and is clear and manifest.[24]
Pre-emption itself is complex and can take several forms. Sometimes, judicial interpretations vary widely for the same federal law’s interaction with different state laws.[25] It may be ‘express’, where Congress explicitly states its intent to pre-empt state law in a federal statute, and courts will look for clear language indicating this intent. Courts can also locate this intent in federal agency actions flowing from “congressionally delegated authority”.[26]
Alternatively, pre-emption may be ‘implict’, where even if Congress does not explicitly state its intent, courts may infer pre-emption. Implicit pre-emption can occur in two ways: (1) field pre-emption, where federal regulation is so comprehensive that it occupies the entire regulatory field, leaving no room for states to supplement federal regulation, or (2) conflict pre-emption, when compliance with both federal and state laws is impossible or when state laws obstruct or frustrate the objectives and purposes of federal law.[27] In cases of implicit pre-emption, the structure and purpose of the federal law must demonstrate that pre-emption is necessary.
With these considerations in mind, we can examine two potential outcomes under federal pre-emption for AI regulation. We list them in no particular order.
A. Express pre-emption on AI. Congress can pass laws that explicitly restrict states from passing laws in the same domain, e.g., a federal law that contains express provisions preventing states from imposing restrictions on the distribution of AI models, particularly open-source models. Thus, if a state were to propose a law requiring developers to obtain pre-approval from a state agency before releasing specific AI models—like California has tried to do—federal pre-emption could override such requirements. In some cases, however, states can propose higher standards than the federal baseline for enhanced safety or to tackle local concerns. The interaction between federal and state approvals would then depend on the contours of such legislation. Mere negation in the absence of any coherent federal policy may lead to a vacuum regarding who can assess the suitability of the models and may not pass judicial review. This will also lead to uncertainties for businesses. The Federal Government could also justify such regulation from the Commerce Clause standpoint to ensure that AI models can freely operate and be distributed across state lines. At the same time, such a federal law will have to articulate a clear policy for governing AI model distribution to satisfy courts.
Express pre-emption strategies have been used by the Federal Government in the field of tech regulation before. For instance, the Internet Tax Freedom Act (“ITFA”) was enacted by Congress in 1998 to prevent state and local governments from charging taxes on “internet commerce” as well as from “multiple and discriminatory taxes on electronic commerce”.[28] This controversial law was passed when federal policy prioritized unhindered the development of Internet services.[29] States have argued that the ITFA infringes on state sovereignty and that e-commerce no longer needs heightened protection. According to experts, the “imprecise” drafting of the ITFA has led technology companies to “creatively” argue for pre-emption when they disagree with state law. For instance, Apple challenged a Chicago law levying taxes on streaming video services, stating that the service would be covered by the definition of “internet access” under the ITFA.[30]
Additionally, as pointed out above, while considering pre-emptive strategies, the Federal Government may have to consider a more substantial intervention. It could require direct oversight of AI model development, including rules around training, safety evaluations, licensing requirements, or liability placed on developers themselves. This would broadly overlap with state-level regulations and restrict states from implementing their own regulatory, licensing, and monitoring frameworks, which deviate from the federal stand. Federal agencies with centralized expertise, such as the National Institute of Standards and Technology (NIST) or another designated body, could be tasked with adopting and enforcing these standards. In this case, some interesting and novel questions of pre-emption in the field of AI can arise. Does the federal AI legislation (general or sectoral) belong to a field typically regulated by the federal government? And is there a conflict between state and federal law, and would this conflict complicate compliance with both laws? Here, existing pre-emption jurisprudence in the field of immigration may be helpful.[31]
B. Implied pre-emption on AI. Even without explicit statutory language, the federal government could establish a broad yet minimal AI regulatory framework (e.g., general guidelines on transparency, safety, or ethics). Courts would then determine whether this implicitly occupies the entire regulatory field (field pre-emption), invalidating state laws, or creates conflicts with state laws (conflict pre-emption). Even a lenient framework could implicitly signal that stricter state regulations are unnecessary or inappropriate.
An example of field pre-emption would be if the federal government delegates model AI standards to federal agencies without expressly barring states. Such an approach would prevent states from passing domain-specific regulations, thereby stifling localized innovation and experimentation. States that pioneered AI regulations (e.g., biometric privacy laws) would lose incentives to continue regulatory innovation. Yet, such a sweeping pre-emption of a general-purpose technology like AI would be challenging to implement and politically contentious, facing significant constitutional challenges.
Conflict pre-emption could arise if federal rules (e.g., by the Department of Commerce or the Federal Communications Commission), even if permissive, (are strategically designed to) clash with stricter state laws. For instance, if a federal regulation grants companies broad rights to collect or use data for AI training or development, any more stringent state-level privacy or AI ethics laws limiting data use could become unenforceable due to conflict pre-emption. Courts might interpret that the state’s more restrictive rules obstruct or frustrate the objective of national uniformity or innovation promotion set forth by federal policy.
Against these two possibilities, a deregulatory administration could deliberately craft broad, high-level principles for AI governance without burdensome specifics. Even a sparse, minimal federal framework—if carefully structured—could imply that federal authorities have comprehensively occupied AI regulation, leaving no room for stricter state-level intervention. Courts may then interpret these minimal regulations as exhaustive, pushing out more rigorous state standards. On top of this, companies and industry groups—aligned with the federal administration’s stance—could actively challenge stricter state AI regulations in federal court, arguing these laws conflict implicitly with federal interests or regulatory objectives.
Notably, the federal government can pre-empt state law without really passing legislation to that effect. In certain circumstances, congressional inaction—particularly when it reflects a deliberate decision not to regulate a specific area—may be sufficient for courts to infer pre-emption.[32] This is called negative pre-emption. In these cases, courts will have to examine the policy and legislative contexts of the particular field to see if Congress “considered, but did not enact detailed regulations”.[33], [34] However, courts carefully apply this principle, and there is greater scrutiny if the legislative field is a “traditional state interest”.[35] For instance, courts will examine whether federal agencies or regulators have clearly expressed that regulation is not merited or appropriate in the particular field. It is unclear how federal agencies will make this claim in particular AI sectors where user harm and societal risks are already significant.
3. Discussion
In the previous analysis, we have outlined the two possible scenarios concerning AI regulation in the US and analyzed some of their policy implications. Predicting which scenario will ultimately prevail is challenging. Nevertheless, several statements in the EO and its accompanying fact sheets could be interpreted as a preference of the Trump Administration towards federal dominance, with some version of a nationwide common approach over a patchwork of state laws. The EO calls for rescinding or revising policies deemed “inconsistent with enhancing America’s leadership in AI”. Moreover, OpenAI’s chief global affairs officer has disclosed that their White House memo explicitly recommends pre-empting state laws to secure AI leadership with democratic values and counter China.[36]
Much of the reasoning favoring deregulation is centered on innovation and ease of business, and a rights-based approach is conspicuous in its absence. It remains to be seen if these concerns are addressed subsequently. As of now, if states were to adopt mandates grounded in safety or human rights, they would be perceived as conflicting with the administration’s pro-innovation agenda—such as stringent transparency requirements or robust ethical constraints. The fact sheet is silent on state interventions in formulating local AI policy. Instead, it emphasizes a robust federal strategy to “eliminate harmful Biden Administration AI policies” and lift “onerous and unnecessary government controls”, suggesting a clear preference for uniform regulations and a light-touch approach.
In this context, the administration may view state-by-state rules as barriers to its pro-innovation agenda and invoke the Commerce Clause, consistent with the EO’s emphasis on maintaining AI “free from ideological bias”, which can vary across states. This inclination toward uniformity is further supported by a report of the Senate’s bipartisan working group on AI, which recently released a roadmap calling for $32 billion in federal funding for AI research, as well as for the development of a framework to guide pre-deployment evaluations of AI models and to set case-specific standards for transparency and explainability.[37] Notably, the roadmap emphasizes enforcing existing laws to address AI-related issues rather than creating a new, model-based regulatory regime similar to those proposed by some states.The contours of such legislative retrofitting remain vague and will likely be formalized by federal agencies soon.
It will be interesting to see whether existing federal agencies are given the mandate of general or sectoral AI regulation or if new institutions are created through law. It should be noted that the Federal Communication Commission’s technology oversight already occupies a contentious position in pre-emption jurisprudence. In 2018, for instance, it passed an Order about broadband Internet access services “expressly” pre-empting “any state or local requirements that are inconsistent with [its] deregulatory approach”.[38] This Order included a pre-emption directive to bar states from enacting their net neutrality and adopt a “uniform set of federal regulations”.[39] However, the Court vacated the pre-emption directive. In particular, it observed that the FCC sought to give the directive “independent and far-reaching” effects,[40] but lacked the authority to do so. It emphasized that the FCC’s power is limited to that which is “congressionally delegated”. This limitation was especially significant because the FCC had classified broadband as a Title I “information service”, which does not confer comprehensive regulatory authority. That level of oversight is only available under Title II, which governs “telecommunications services”. Therefore, the court noted that for a federal agency to pre-empt state law lawfully, it must possess either “express” statutory authority or valid “ancillary” authority—neither of which the FCC demonstrated in this case.[41]
In areas traditionally regulated by states—such as health, safety, and welfare—courts often presume that Congress did not intend to pre-empt state law unless there is clear evidence to the contrary. At the same time, historical precedents in technology sectors—like autonomous vehicles, medical devices, telecommunications, and smartphones—suggest that federal pre-emption is plausible for AI. In these cases, the federal government has often targeted general-purpose technologies—those with broad applications across industries and daily life—to streamline production and distribution. Just like electrical outlets and cellular connectivity, which adhere to consistent technical standards nationwide, AI may similarly warrant a unified federal approach.[42] The longstanding debate on consistency between state and federal law in the field of privacy and data protection can also be instructive.[43]
Currently, many states have enforced consumer data protection laws far stronger than those in proposed federal laws, such as the American Data Privacy Protection Act. In such scenarios, experts have pointed out how federal privacy laws should be seen as “floors” and not “ceilings”,[44] i.e., federal standards should not water down higher privacy protections that states may offer. Instead, they should set a baseline that can be contextually topped up by states but outlines reasonable rights and duties.[45] Though not yet adopted in data privacy, this approach may be helpful for federal AI law-making. However, as said, the possibility of a federal law grounded in rights remains uncertain.
There is a final consideration. Pre-emption is neither straightforward nor automatic. Courts will first attempt to interpret federal law in a way that allows state law to operate concurrently, avoiding conflicts where possible. Several legal and procedural hurdles could complicate efforts to impose uniform federal standards over state-level regulations. For example, the Administrative Procedure Act (APA) requires exhaustive justifications for federal agency actions that may pre-empt state laws, adding a layer of scrutiny to such decisions. Courts have also applied the “impossibility exception” in limited scenarios, i.e., federal agency authority to pre-empt is valid if (i) the matter has “both interstate and interstate aspects”, (ii) it fulfills a “valid federal regulatory objective”, and (iii) state law would undermine the federal agency’s authority because “regulation of interstate aspects cannot be unbundled from the regulation of the intrastate aspects”.[46]
Finally, in the field of technology regulation, courts have acknowledged how “the realities of technology and economics” make it difficult to assess a federal agency’s power to regulate[47] and give rise to “jurisdictional tensions”.[48] Yet, one of the most significant variables for our discussion is that, while, as mentioned, technological fields have historically seen federal pre-emption, AI applications often intersect with domains traditionally regulated by states, such as health, education, or welfare. The interpretative standard of presumption against pre-emption in these areas is robust, as states have historically exercised primary regulatory authority. As a result, the Supreme Court may be reluctant to endorse broad federal pre-emption without explicit congressional intent. In cases of conflict pre-emption to apply, courts usually consider whether the state law would hinder “the full purposes and objectives of Congress”[49] and whether the lack of Congressional consensus on AI policy so far will complicate matters further.
4. Conclusion
There are compelling reasons to believe the Trump administration is interested in pursuing some form of federal dominance. We focused on pre-emption, examining its implications and challenges.
On the one hand, federal pre-emption would align with the Trump Administration’s broader policy agenda, which prioritizes deregulation and innovation, being at odds with many state-level AI regulations. Indeed, the likelihood of such federal intervention increases as states adopt more stringent rules. As states adopt increasingly restrictive AI laws, federal policymakers or private litigants could argue that such laws conflict with federal objectives, giving rise to creating conflict pre-emption. Historical precedents in the technology sector further support this scenario, highlighting previous instances where pre-emption has been effectively deployed.
On the other hand, the general-purpose nature of AI heavily relies on application-specific regulation, suggesting that states will retain regulatory authority in domains such as health, education, and welfare—areas where states have historically exercised jurisdiction. In this context, federal authority would likely focus on general aspects of AI regulation relevant to interstate commerce, national security[50], and economic competitiveness.
Predicting whether and how federal dominance will materialize remains challenging. Uncertainty persists regarding the viability of general or sector-specific AI legislation in a divided Congress, particularly given fluid support and lobbying from the technology industry. Given AI’s general-purpose characteristics, it appears more appropriate for sector-specific regulations, preserving state authority over traditionally state-controlled areas while reserving for the federal government those areas requiring national coordination, such as commerce, security, and global economic positioning.
Ultimately, considering AI’s legal and technical complexities, the current administration’s push toward deregulation will likely need some form of federal intervention. As suggested by the intricacy of pre-emption jurisprudence, achieving these deregulatory goals at the federal level may require substantive and targeted legislative or policy measures.
[1] The new Executive Order can be found here: https://www.whitehouse.gov/presidential-actions/2025/01/removing-barriers-to-american-leadership-in-artificial-intelligence/. Meanwhile, President Biden’s Executive Order No. 14110, dated October 30, 2023, is titled “Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence”.
[2] EO n. 13859 of February 2019 (“Maintaining American Leadership in Artificial Intelligence”) and EO n. 13960 of December 2020 (“Promoting the Use of Trustworthy Artificial Intelligence in the Federal Government”).
[3] While these EOs mentioned the importance of civil liberties, they did not include any concrete enforcement mechanisms or oversight measures to protect them.
[5] More specifically, under the current executive order, within 180 days, the Assistant to the President for Science and Technology (APST), the Special Advisor for AI and Crypto, and the Assistant to the President for National Security Affairs (APNSA)—in coordination with other advisors and agency heads must develop and present an AI action plan to the President.
[6] https://www.lemonde.fr/en/united-states/article/2025/02/11/jd-vance-rails-against-excessive-regulation-of-ai-at-paris-summit_6738021_133.html
[7] [Trump E.O. cited above]
[8] https://www.dataguidance.com/opinion/usa-common-themes-2024-state-ai-legislation-and
[9] The initiative tasked the National Institute of Standards and Technology (NIST) with developing technical standards for trustworthy and secure AI systems.
[11] Two fact sheets may be significant for our analysis: 1) “President Donald J. Trump Takes Action To Enhance America’s Ai Leadership” And 2) “Executive Order To Establish United States Leadership In Digital Financial Technology”.
[12] [Trump EO cited above]
[14] https://www.cbia.com/news/issues-policies/sweeping-artificial-intelligence-bill-stalls/.
[15] You can read the Act here: https://publications.tnsosfiles.com/acts/113/pub/pc0588.pdf.
[16] See the updated database of the National Conference of State Legislatures here: https://www.ncsl.org/technology-and-communication/artificial-intelligence-2025-legislation
[17] [to add reference for data]
[18] California Senate Bill (SB 1047). https://thehill.com/opinion/technology/4739275-preemption-ai-regulation/.
[19] This point has been also stressed by Dean W. Ball and Alan Z. Rozenshtein in https://www.lawfaremedia.org/article/congress-should-preempt-state-ai-safety-legislation.
[20] Mira Moldawer, “ELVIS Act: From Authorship to Ownership in Intellectual Property Law,” Texas Intellectual Property Law Journal 33, no. 1 (2024): 19-66; First Amendment challenges are also foreseen for other AI laws related to campaign speech such as H.R. 3044 and S. 1596 which seek to amend the Federal Election Campaign Act , Artificial Intelligence in Federal Election Campaigns: Legal Background and Constitutional Considerations for Legislation
[21] Some bipartisan laws introduced in the last few years indicate a possible movement on sectoral legislation at the federal level. E.g. the TAKE IT DOWN Act, https://www.commerce.senate.gov/2024/7/sen-cruz-s-take-it-down-act-clears-commerce-committee
[22] Stephen A. Gardbaum, “Nature of Preemption”, Cornell Law Review 79, no. 4 (1993-1994): 767-815; Nelson, Caleb. “Preemption”. Virginia Law Review, vol. 86, no. 2, 2000, pp. 225–305. JSTOR, https://doi.org/10.2307/1073916. Accessed 25 Jan. 2025.
[23] Rice v. Santa Fe Elevator Corp. (1947).
[24] A presumption in favor of state sovereignty is adopted by courts in the “anti-commandeering” jurisprudence as well. Here, the federal government cannot “commandeer” states to “carry out its will”. https://journals.law.harvard.edu/crcl/wp-content/uploads/sites/80/2020/10/Butash.pdf.
[25] https://scholarship.law.missouri.edu/mlr/vol58/iss2/3
[26] Louisiana Public Service Commission v. FCC
1947).
aum, ibidem, pp. 770 ff.
[28] https://crsreports.congress.gov/product/pdf/IF/IF11947
[29] https://digitalcommons.law.uga.edu/cgi/viewcontent.cgi?article=2523&context=fac_artchop
[Apple Inc. v. City of Chicago], [see also Appe Inc. v. Hegar]
[30] Apple Inc. v. City of Chicago
[31] For instance, In De Canas v. Bianca, the Supreme Court laid out three tests to determine federal pre-emption for immigration statutes.
[32] https://scholarship.law.missouri.edu/mlr/vol58/iss2/3/, p. 4.
[33] ibidem.
[34] Northfolk & W. Ry. v. Public Utils. Comm’n of Ohio, 926 F.2d 567, 570 (6th Cir. 1991).
[35] ibidem.
[36] https://www.axios.com/2025/03/13/openai-chris-lehane-trump-policy.
[37]You can read a summary here:https://www.schumer.senate.gov/imo/media/doc/LD%20One_Pager_Roadmap_DRAFT_Clean_MK_YD_lD.pdf.
[38] Mozilla Corp v. FCC, No. 18-1051 (D.C. Cir. 2019).
[39] Id.
[40] Id, p. 122.
[41] Id.
[42] https://www.lawfaremedia.org/article/congress-should-preempt-state-ai-safety-legislation.
[43] For instance, Colorado’s SB 25 is a comprehensive AI legislation which impacts AI operations in healthcare, government, education, finance, employment etc.
[44] https://www.eff.org/deeplinks/2022/07/federal-preemption-state-privacy-law-hurts-everyone, https://iapp.org/news/a/ceiling-or-floor-state-law-preemption-and-preservation-in-u-s-federal-privacy-bills
[45] https://www.eff.org/deeplinks/2022/07/federal-preemption-state-privacy-law-hurts-everyone
[46] Maryland PSC, 909 F.2d at 1515 [p.128, Mozilla v. FCC] “impossibility exception does not create pre-emption authority out of thin air” p. 129 Mozilla v. FCC
[47] Louisiana PSC, 476, U.S. at 360
[48] P. 128, Mozilla v. FCC
[49] P. 136, Mozilla v. FCC
Luciano Floridi
Luciano Floridi (Laurea, Rome La Sapienza, M.Phil. and Ph.D. University of Warwick) is the John K. Castle Professor in the Practice of Cognitive Science and the Founding Director of the Digital Ethics Center. His areas of research are the philosophy of information, digital ethics, the ethics of AI, and the philosophy of technology.
Akriti Gaur
Akriti Gaur is an Indian lawyer currently pursuing a J.S.D. at Yale Law School where she served as a Tutor in Law from 2022-2024. She obtained her LL.M. degree from YLS in 2022.
