Author: Revaz Topuria, Research Fellow
On June 2, 2026, President Trump signed the executive order on “Promoting Advanced Artificial Intelligence Innovation and Security” (our analysis of this order is available here). Two days after Trump’s executive order, two members of Congress, from both parties, released a legislative draft for governing artificial intelligence. Congressman Jay Obernolte, a California Republican, and Congresswoman Lori Trahan, a Massachusetts Democrat, introduced a 269-page discussion draft titled the “Great American Artificial Intelligence Act of 2026.” If President Trump’s executive order demonstrated the administration’s readiness to regulate the AI sphere without Congress, this bill attempts to bring AI governance into a clearer legislative framework.
The introduction of this bill is a fairly significant step in American AI policy. A presidential executive order reflects one administration’s preferences and can be reversed by the next administration. A law, once passed, is far more durable, binding on future administrations as well, and represents the product of mutual agreement between both parties. The “Great American AI Act” is the first real test of whether Washington can resolve various AI-related questions, whether protecting innovation or addressing risks, through legislative agreement between the two poles.
What does the bill cover?
The bill is divided into four parts. The first and most significant chapter concerns the leading AI models themselves. Companies with annual revenue exceeding $500 million will be required to disclose information about how their systems work, and to take steps to mitigate certain risks before releasing a new model. The bill also gives permanent legal standing to the Center for AI Standards and Innovation, an internal agency of the Department of Commerce.
The second part addresses the AI workforce and includes funding for retraining programs for people who lose their jobs due to automation. It also covers introducing AI literacy and expanding AI education from kindergarten through university.
The third part strengthens cybersecurity, extending until 2035 the legal protections that currently shield companies from antitrust liability when they share information about cyber threats with one another.
The fourth part funds AI research and development and determines how the United States should coordinate with its allies on AI policy issues at the international level. This is precisely the area that connects directly to competition with China. This cooperation primarily takes place in bodies such as the UN, the G7, and the Organisation for Economic Co-operation and Development, where Washington is trying to unite its allies around AI standards based on democratic values. At the same time, China is actively developing its own alternative, active participation in international standards bodies and launching initiatives such as the World AI Cooperation Organization, thereby drawing the Global South toward its own governance model.
For its part, most of the bill is pretty straightforward; however, one provision in the section on leading AI models has generated considerable debate. Under this rule, individual states would be barred for three years from passing their own laws related to the creation of AI models. States would retain full authority over how AI is used within their borders, so a state would still be able to regulate companion chatbots, various algorithms, or facial recognition technologies. What they would lose for the duration of the moratorium is the power to set their own rules for how companies train and test leading models.
Federal AI Policy
The debate over which branch of government should regulate what, or whether it should be regulated at all, whether the approach should be federal or whether states should have their own autonomy, is certainly nothing new in American politics. Nor is AI policy an exception. The central argument of Washington’s July 2025 AI Action Plan and Trump’s executive order is that a single federal standard serves American competitiveness better than fifty different state laws, each with its own distinct licensing requirements and different standards. Large American technology companies have argued for years that complying with dozens of different regulations and standards across various states slows them down so much that they lose competitive advantage against Chinese companies, which operate under only one state regulatory body. According to the bill’s authors, the goal of the Great American AI Act is precisely to give this argument the force of law.
The Bill’s Uncertain Future
It is clear that this bill has gained plenty of supporters as well as opponents. If on one side the argument speaks to an effective, centralized legislative framework, the bill’s opponents see the risks of crude federal government interference and diminished state autonomy. Particularly problematic, as we noted, is the three-year moratorium being considered on states passing new AI laws.
This discussion has brought to light an even deeper issue, viewed through the lens of US-China competition in AI. China’s government does not have to hold any similar discussion publicly at all. When Beijing decided to regulate generative AI services, require content labeling, or incorporate AI oversight into cybersecurity law, it issued administrative acts through the Cyberspace Administration and other ministries, and these mechanisms simply took effect, without public or parliamentary debate. In China’s case, there was no 269-page discussion draft that spent months gathering feedback from interested parties, nor was there any risk that a rival faction within the same ruling party would publicly declare the project “dead on arrival” as soon as it was published, as happened in the American case. The Chinese system trades transparency and pluralism for speed, whereas in America’s case, even after months of discussion, the “Great American AI Act” remains a discussion draft rather than a law, with no guaranteed date for formal introduction.
However, it is clear that this approach also has its own distinct advantages. A law that survives this kind of scrutiny, that takes into account the views of two opposing parties as well as the opinions of industry, labor unions, and civil society groups, will acquire a solid legitimacy that no executive order can match, and it will not be simple for the next occupant of the White House to overturn it with a single signature.
For three years, American AI policy has consisted almost entirely of a mixture of executive orders, various agency guidelines, and state laws, without a single federal law that would tie these pieces together. The “Great American AI Act” is the first serious attempt to change this, and at the same time it is a test of how well the United States can create a unified legislative framework to consolidate its own leadership in AI research and investment.
This article was translated from the original language with the assistance of AI tools and revised by the author.