[co-author: Ted Loud]
Provision in House-passed “reconciliation” bill would bar states and localities from enforcing laws or regulations on AI models
With an eye on Europe’s heavy regulation of AI Systems, the proliferation of US state bills and attorney general actions looking to localize AI governance, and largely competing views on AI safety, the US Congress is now entertaining a new steady state: Overriding the state patchwork of AI laws while carving out a space for potential – but not guaranteed – federal control.
On May 22, 2025, by a vote of 215-214, the US House of Representatives passed a set of significant AI-related provisions in the budget reconciliation package currently making its way through Congress. The provisions include funding for government technology upgrades incorporating commercially developed AI and a ban on state and local enforcement of AI legislation and regulations.
AI modernization
Subtitle C, Part 2, titled “Artificial Intelligence and Information Technology Modernization,” would appropriate $500 million to the Department of Commerce, to remain available through fiscal year 2035, to modernize and secure federal information technology systems through the deployment of commercial AI, automation technologies, and the replacement of antiquated business systems.
The Secretary of Commerce is directed to:
use these funds to support the replacement and modernization of legacy business systems with state-of-the-art commercial artificial intelligence systems and automated decision systems, the adoption of artificial intelligence models that increase operational efficiency and service delivery, and improve the cybersecurity posture of Federal information technology systems through modernized architecture, automated threat detection, and integrated artificial intelligence solutions.
The budgetary allocation follows a series of government initiatives designed to bring government technology and use of AI into the modern era, including the White House Office of Management and Budget’s recent publications that include guidance on the acquisition and use of AI in government in alignment with the Trump Administration’s broader AI strategy.
AI moratorium
Subsection (c), titled “Moratorium,” states that, subject to limited exceptions, “no state or political subdivision may enforce, during the 10-year period beginning on the date of the enactment of [the moratorium] any law or regulation limiting, restricting, or otherwise regulating artificial intelligence models, artificial intelligence systems, or automated decision systems entered into interstate commerce.”
The prohibition relies on several defined terms that depart from interpretations currently favored by international parties and the Biden Administration. These include the following:
- “Artificial Intelligence” maintains the definition included in section 5002 of the National Artificial Intelligence Initiative Act of 2020 (15 U.S.C. 9401): “a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments. Artificial intelligence systems use machine and human-based inputs to-
- (A) perceive real and virtual environments;
- (B) abstract such perceptions into models through analysis in an automated manner; and
- (C) use model inference to formulate options for information or action.”
- “Artificial Intelligence Model” is defined as “a software component of an information system that implements artificial intelligence technology and uses computational, statistical, or machine-learning techniques to produce outputs from a defined set of inputs.”
- “Artificial Intelligence System” is defined as “any data system, hardware, tool, or utility that operates, in whole or in part, using artificial intelligence.”
- “Automated Decision System” is defined as “any computational process derived from machine learning, statistical modeling, data analytics, or artificial intelligence that issues a simplified output, including a score, classification, or recommendation, to materially influence or replace human decision making.”
The interaction of these definitions with others in use – all complex – will likely give rise to additional intricacies in navigating the global patchwork. Construction of the moratorium has already raised other questions regarding its interpretation. For example, whether potential liability under state or local government AI legislation begins after ten years, or whether states will be able to look back into the ten-year period to investigate potential breaches, has been considered unclear.
Clarity on these questions could help determine the moratorium’s potential impact on private litigation or state enforcement of existing statutes of general applicability (ie, those that do not single out AI but nevertheless regulate its use, such as consumer protection legislation).
Legislative and procedural challenges
It remains to be seen whether the proposed provisions, particularly the moratorium on AI regulation, will survive the legislative process intact – or at all. While the text did pass on the floor of the US House of Representatives and overcame an amendment to strike the moratorium language during the initial Energy and Commerce Committee markup on May 14, 2025, the text continues to face ongoing challenges.
Committee challenges
During the committee meeting, Ranking Member Frank Pallone proposed to strike the moratorium, which he called a “giveaway to big tech.” Several Democratic members spoke in support of Pallone’s amendment, with members also saying that state laws provide an important backstop, at least until Congress can enact meaningful legislation. Some Democratic members also stated that states’ attempts to regulate AI could help inform federal action, and expressed skepticism of Republican representatives’ desire to take federal action.
On the Republican side of the committee, Representative Jay Obernolte spoke against the Pallone amendment, pointing to the difficulty that entrepreneurs and private companies would face in navigating 50 different state legislative environments. Representative Obernolte stated that the Constitution gives Congress power over interstate commerce and that Congress should take action similar to that of the federal framework that the House AI Task Force, which he co-chaired, crafted last year. Representative Obernolte expressed hope that the moratorium would not need to remain in place for the full ten years.
Committee Chair Brett Guthrie echoed Representative Obernolte and stated that a patchwork of state laws would be unworkable given the interstate factors AI companies face, such as data centers that source their power from another state.
Representative Laurel Lee also spoke in support of the moratorium, stating that, while big companies could have the resources to navigate a patchwork of laws, small innovative companies would be unable.
External criticisms
Critics of the moratorium, including a group of state lawmakers and AI researchers, have argued that the moratorium would prevent states from being able to enforce existing key AI regulations, including those prohibiting deepfakes in election campaigns and safeguarding against automated decision-making. Such positions include the argument that a moratorium of this nature would be in contravention of the 10th Amendment on the basis that Congress does not have the power to act in this way. Others have argued that the moratorium does not relate to budgetary matters and should be removed, leaving the provisions providing for funding to enable AI modernization in government.
Federal preemption
The question of whether federal AI legislation and regulations should preempt initiatives at the state level has been an ongoing subject of debate in Congress, which has thus far been unable to enact a national framework for how AI is policed. Given the slow pace of congressional action, many state legislators have been eager to fill the void and push their own solutions. Many in Congress, as well as the governors of California and Colorado, have argued that a state-by-state approach would create a patchwork or piecemeal regulatory regime that could create confusion over enforcement and compliance for industry innovators, justifying a need for federal control. Federal preemption of this nature (where there is currently no federal legislation on the subject) is uncommon. Distinct federal AI legislation would not be required to establish preemption of state-law regulation; however, the moratorium’s language in the proposed budgetary bill would leave some uncertainty about the scope of such preemption.
The Senate "Byrd Rule"
Now that the House of Representatives has passed its version of the reconciliation bill on May 22 that includes the proposed ten-year moratorium on AI regulation, the focus turns to the Senate, which faces notable procedural hurdles. Budget reconciliation is an expedited parliamentary procedure that prevents filibusters and limits amendments in the Senate. Reconciliation bills are meant to strictly address spending, revenue, and the federal debt limit. The Byrd Rule, named for the late Senator Robert Byrd (D-WV), is a provision of the Congressional Budget Act that restricts the inclusion of non-budgetary “extraneous” policy provisions in reconciliation bills. Any senator may raise a point of order against a provision for violation of the Byrd Rule, which is ruled on by the presiding officer of the Senate based on the advice of the Senate Parliamentarian. If the Parliamentarian deems the provision to be out of order, that ruling can be waived by a 60-vote supermajority of senators, which may be unlikely given the 53-seat Republican majority. The Vice President (in his capacity as President of the Senate) can also overrule the parliamentarian, but such overrule has not happened in 50 years.
During the Energy and Commerce markup, as some Democratic representatives sought to portray the moratorium as a policy change that would violate the Byrd Rule, Representative Obernolte defended the provision’s suitability for inclusion in a budget reconciliation bill. He stated, “My understanding is that the purpose of this provision is to spend $500 million to modernize federal IT and cybersecurity systems. To be clear, the moratorium is a necessary term and condition to execute the primary purpose of the provision.”
Some Senate Republicans, including Senators Marsha Blackburn (R-TN) and Josh Hawley (R-MO), have voiced concern over the proposed moratorium.
Even if the moratorium language is removed from the budget reconciliation because of the Byrd Rule, some congressional Republicans have signaled their intention to pursue a moratorium by other legislative means. On May 15, 2025, Senator Ted Cruz (R-TX) announced his plans to include a ten-year moratorium on state and local AI laws and regulations as part of a larger AI policy bill he intends to introduce in the near future. Senator Cruz chairs the Senate Commerce Committee, which has jurisdiction over communications and technology issues.
For innovators and adopters, the effort to reduce complexity in AI regulation could raise as many questions as it answers, adding new definitions and debates about federal/state and US/global interactions. But, with the extraterritorial impact of the EU AI Act, as well as the application of longstanding legal principles from tort and contract litigation to government enforcement of laws of general applicability, the throughline for companies remains the same: Responsible AI (RAI) is smart AI, and RAI could drive return on investment.
[View source.]
