Menu
This past summer has seen major updates in the burgeoning relationship between the U.S. government and developers of the latest artificial intelligence (“AI”) models. On June 2, 2026, President Trump signed an executive order, entitled “Promoting Advanced Artificial Intelligence Innovation and Security” (the “EO”).1 And on June 4, 2026, Congressional Representatives Lori Trahan (D-MA) and Jay Obernolte (R-CA) released a draft bill styled as “the Great American AI Act,” (the “GAAIA”) and are soliciting comments before introducing this bill to Congress.2
Prior to the June 2 EO, the Trump administration had used a laissez-faire approach with regard to AI.3 But the timing of this EO coincides with other developments in the AI industry, and it may signal a shift in the administration’s policy toward AI regulation.
The EO came amid important legal and commercial developments affecting AI. Just before the EO, on June 1, Florida sued OpenAI for public nuisance and for maintaining a dangerous online product.4 Meanwhile, OpenAI and Anthropic, two of the largest AI developers, have been pursuing their initial public offerings.5
This EO, issued at a pivotal moment before public listing of some of the largest AI developers, may signal that the administration intends to regulate AI more stringently moving forward. As things stand, the White House attributes much of the success of the AI industry in the U.S. to their “refus[al] to stifle this innovation with overly burdensome regulation.”6 But policies aimed at promoting AI innovation must also contend with the public concern that AI development proceeds without the necessary safeguards.7 Both the EO and the GAAIA seek to balance innovation and regulation amid the constant progress and cybersecurity implications of advanced AI models.8
The EO affirms the White House’s aim to encourage AI innovation, while promoting government oversight of AI development to address national security considerations.9 To these ends, the EO proposes to create a partnership between government agencies and AI developers to confront the emerging threats that come with deploying newer and more powerful AI models.
Specifically, the EO requires that the Secretary of the Treasury establish “an AI cybersecurity clearinghouse in voluntary collaboration with the AI industry and operators of critical infrastructure.”10 This clearinghouse would promote coordination between the government and private AI developers to identify software vulnerabilities, remediate them, and distribute patches.11
The EO also requires government agencies to develop and maintain a classified benchmarking process to assess the capabilities of AI models and determine when an AI model will be considered a “covered frontier model.”12 The government will use this benchmarking process to determine which AI models are “covered frontier models” and the EO creates a voluntary framework wherein AI developers of these models can engage with the government, providing access to these models for 30 days ahead of release.13 Providing advance access to covered frontier models could allow for meaningful review of the security vulnerabilities revealed by these models before they are released to the public.14
But the long-term effect of the EO may be limited because participation is voluntary. As the White House put it in a social media post, “[w]e are NOT conducting oversight of all new models, as that level of government overreach would have chilling effects on free speech and innovation.”15 Chilling effects or no, it may also be that the President simply lacks power to compel participation in this framework without new legislation.
But voluntary adherence may be hard to get. AI developers may not see any reason to pause their progress pending the 30-day government review, particularly in view of the frenzied competition for market share. And while the EO’s 30-day review window may seem to be too long because of the demands of the marketplace, it may also be too short for adequate pre-deployment testing.
Additionally, deciding what is a “covered frontier model” may hinder even applying the new framework in the first place, as AI developers and government reviewers may not agree in their definitions. Considering these hurdles, the EO may have little effect on AI development unless developers see sufficient countervailing advantages to participating.16
The EO generated mixed reactions from AI-industry insiders. Some AI executives are reported to have praised the EO as an important step to balance AI safety and innovation.17
But some large developers have reservations. OpenAI, for example, called for a mandatory evaluation system, saying that “the world needs more than voluntary commitments . . . [i]t needs harmonized legal frameworks and durable institutions.”18 Anthropic went further, calling for a global slow-down or even pause in AI development.19 Anthropic said that “[i]f it were possible to effectively slow the development of this technology to give ourselves more time to deal with its immense implications, we think that would likely be a good thing.”20
Some of Anthropic’s concerns stem from the possibility for “recursive self-improvement,” when AI models develop and improve other AI models, circumventing or overwhelming human oversight.21 Likening the need to control the AI-race to a ballistic missile disarmament treaty, Anthropic urged for better coordination and deliberation before deploying the latest models.22 “Without a global coordination mechanism, companies and governments will have to make difficult decisions about safety while under competitive and geopolitical pressures.”23
Another development on the AI landscape is the draft of the “Great American AI Act” (“GAAIA”) that was circulated for comments earlier this summer.24 Representatives Trahan and Obernolte, both members of the House Energy and Commerce Committee, published this draft legislation along with a companion op-ed on Bloomberg Law titled “America Needs One National Framework for Artificial Intelligence.”25 The draft GAAIA seeks to create a federal framework that would govern AI development and position the U.S. as the global standard-setter for vetting AI models.26
Under the GAAIA, large frontier AI developers (defined as those with > $500 million in revenue), would be required to publish and implement catastrophic risk plans before deploying new AI models.27 The GAAIA would also create the Center for AI Standards and Innovation (“CAISI”), a government agency that would promulgate AI development guidelines and standards with rulemaking authority.28 AI developers would be required to report safety incidents and submit to audits overseen by CAISI.29
The GAAIA would also require government stakeholders to engage in the development of international AI technical standards and advocate for the adoption of U.S.-developed AI standards.30 In their own words, the GAAIA sponsors hope that this framework may strengthen the U.S.’s ability to lead internationally, and will “help ensure AI develops in a way that reflects democratic values, economic freedom, and the rule of law.”31
The parts of the GAAIA that have received the most attention so far, however, are the provisions that would preempt state laws regulating AI model development. The bill’s sponsors assert the need for a federal framework, saying that “risks from AI don’t stop at state lines.”32 Making the federal framework exclusive, they add, is needed to protect innovators from having “to navigate dozens of legal regimes.”33 The effect would be to treat regulation of AI model development similarly to patent law, as a field governed exclusively by federal law.34
There are limits to the proposed preemption of state law, however. The GAAIA distinguishes between laws regulating AI model development, where states are preempted, and laws regulating AI deployment, which states may still enforce.35 Further, the GAAIA does not purport to preempt laws of general applicability or state laws regulating AI use or deployment.36 And states would keep their authority to enact consumer protection laws banning deepfakes, maintaining privacy, or requiring notifications of AI use, for example.
Outside commentators, including the American Civil Liberties Union, have criticized the GAAIA’s preemption provisions, concerned that the federal preemption of state regulations on AI development may cripple vital accountability measures.37 The ACLU contends, particularly, that the states’ retained ability to regulate AI deployment would be illusory because AI models may have violations of state laws encoded into the system during the development stage.38 Deployment may therefore be too late to address violations of state law that are baked in to an AI model during its development.39 The ACLU therefore concludes that the GAAIA would preempt effective regulation that is necessary for safe, responsible, and non-discriminatory AI technology.40
Another consideration is that, although the GAAIA may reduce the complexity of complying with regulatory regimes that vary between U.S. states, it may complicate compliance with varying regimes internationally. For example, many AI developers already comply with the EU AI Act. Similar to the GAAIA, the EU AI Act creates transparency obligations for AI developers and requires “conformity assessment procedure” audits of high-risk AI systems.41 And while there may be overlap between the EU AI Act and the GAAIA, as the GAAIA matures into law, significant differences between the two regimes may emerge and make compliance with both systems difficult for AI developers.
As the GAAIA draft circulates and collects comments, many open questions remain, especially concerning federal preemption and compliance. Depending on the outcomes to these questions, the regulatory environment for AI developers in the coming years may vary drastically.
Innovation in disruptive technologies often runs far ahead of government regulation, and this pattern is repeating itself for AI. As AI models affect more facets of our lives and their reach spreads, there may be more and stronger calls for a national regulatory regime to ensure these models are deployed safely. And besides raising concerns about privacy, intellectual property, and consumer protection, developments in AI may also introduce “new national security considerations” that require government response.42
The June 2 EO may be a useful starting point for engagement between AI developers and government regulators, but the voluntary nature of its framework may limit the EO’s long-term effects. The more consequential debates may be about the “Great American AI Act,” and the balance between freedom of innovation and government control. This balance will continue to shift for AI developers, so a close eye on the progress of this bill will be important.