On June 5, bipartisan members of the U.S. House of Representatives released a 269-page discussion draft on artificial intelligence legislation, aiming to comprehensively incorporate the safety audits of cutting-edge AI models, employee whistleblower protections, and research on AI's impact on employment into a federal framework. The draft, titled "The Great American Artificial Intelligence Act of 2026," was led by Republican Congresswoman Jay Obernolte from California and Democratic Congresswoman Lori Trahan from Massachusetts. It is currently in the feedback stage and has not yet been formally submitted for voting.
The most notable focus of this draft is its strict regulations on "large frontier developers." The bill targets companies with annual revenues exceeding $500 million in the previous year and that have trained large frontier models, including leading labs such as OpenAI, Google DeepMind, and Anthropic. These companies must publicly disclose a "frontier AI framework," detailing how they identify and prevent "catastrophic risks" such as assisting in the creation of mass destruction weapons, launching cyberattacks, or causing more than 50 deaths or over $1 billion in property damage without human supervision. To this end, companies are required to hire independent verification agencies to review their safety plans every six months. Violators may face civil fines of up to $1 million per day.
The most controversial provision in the draft is the "state law preemption." The bill states that for three years after its enactment, state and local governments cannot enact or enforce laws specifically regulating AI model development, although they can still regulate the use and deployment of AI systems. Supporters argue that the U.S. should not be slowed down by 50 different sets of development rules across states; however, opponents worry that this will force states to lower their strict standards on consumer protection and children's safety to the federal minimum.
In addition to focusing on tech giants, the draft also extends its reach into the daily lives of ordinary people. The bill establishes whistleblower protection provisions for AI company employees and contractors to prevent retaliation for reporting violations. In terms of consumer protection, the draft increases sentencing standards for financial fraud using AI and creates a specific crime for AI impersonating federal officials. For students and workers, the bill requires research on AI's impact on employment and promotes AI literacy education at the K-12 level and university scholarship programs. Additionally, the draft plans to extend the Cybersecurity Information Sharing Act until the 2035 fiscal year, continuing to allow companies to share information on cyber threats.
Although this draft has generated some attention within the industry, its political prospects remain uncertain. The industry group NetChoice generally supports its direction but expresses concerns about mandatory audits and data sharing potentially leaking trade secrets. Previously, Congress had attempted to include a 10-year state-level AI legislation moratorium in the budget bill, which was ultimately removed by an overwhelming vote of 99:1 in the Senate. This time, the draft shortens the period to three years and includes transparency and audit requirements. Whether it can gain enough support in the modifications and negotiations before the August congressional recess remains to be seen.
