On September 10th, Senate Commerce, Science and Transportation Chairman Ted Cruz (R-TX) released what he called the “Light Touch” regulatory framework for federal AI law, outlining five pillars to advance American AI leadership. In parallel, Senator Cruz introduced enhanced normalization and spread of AI through the Surveillance and Experimental (“Sandbox”) Act (“Sandbox”) Act (“Sandbox”) that establishes a federal AI-regulated sandbox program that abandons or changes federal agency regulations and guidance for AI developers and deployers. Collectively, the AI ​​Framework and Sandbox Act shows the first Congressional effort to implement recommendations for the AI ​​Action Plan released by the Trump administration, released on July 23rd.
Light Touch AI Regulation Framework
Titled “Legal Framework for American Leadership in Artificial Intelligence,” Sen. Cruz’s AI Framework calls on the United States to “encourage entrepreneurial freedom and a history of innovation” by adopting AI laws that prevent “evil uses” of AI technology and promote innovation. Reflecting President Trump’s January 23rd executive order on “Removing barriers to American leadership in artificial intelligence” and recommendations for the AI ​​Action Plan, the AI ​​framework sets five pillars as “the starting point for discussion.”
Unleashes American innovation and long-term growth. The AI ​​framework recommends that Congress establish a federal AI regulatory sandbox program, provide access to federal datasets for AI training, and streamline AI infrastructure permissions. This pillar reflects the priorities of the AI ​​Action Plan and President Trump’s July 23rd executive order. Protects freedom of speech in the age of AI. In line with President Trump’s July 23 executive order on “preventing awakened AI in the federal government,” Sen. Cruz called on Congress to “stop government censorship” (“Joe Borning”) on AI platforms and to address foreign Americans censorship. Additionally, the AI ​​Action Plan revised the National Institute of Standards and Technology (“NIST”) AI risk management framework to “eliminate misinformation, diversity, equity, and climate change references,” but this pillar calls for NIST to “reform AI priorities and goals.” Prevents the burdensome patchwork of AI regulations. Following the failed attempt by Congressional Republicans to enact moratoriums on enforcing state and local AI regulations in July, the AI ​​Action Plan called on federal agencies to consider restrictions on states with AI-regulatory regimes that are burdened with federal AI-related funds, as well as state AI laws that may be scheduled under the communications laws. Similarly, the AI ​​framework calls on Congress to establish federal standards to counter “overly foreign regulations” for Americans, while also preventing AI regulations in burdensome states. Stop malicious use of AI against Americans. Nodding to the bipartisan support for the protection of digital replicas in the states that ultimately doomed Congress’ AI moratorium this summer, the pillar is calling on Congress to protect Americans from digital impersonation and fraud. Additionally, the pillar calls on Congress to expand the principles of the federal takedown laws, which were signed into law in May, to protect American school children from intimate visual depictions of unconsensuality. Protects human values ​​and dignity. This pillar appears to be expanding the US’s “global AI domination to promote human prosperity,” established by the January 23rd executive order. Sandbox Method
Consistent with the AI ​​Action Plan and AI Framework recommendations, the Sandbox Act directs the White House Science and Technology Policy (“OSTP”) to establish and operate the “AI Regulatory Sandbox Program” with the aim of promoting AI innovation, the development of AI products and services, and the expansion of AI-related economic opportunities and employment. According to a press release from Senator Cruz, the Sandbox Act presents the “first step” in implementing AI action plans that demand “excellence in regulatory sandboxes or AI excellence” that allow researchers, startups and established companies to quickly deploy and test AI tools.”
Program application. The AI ​​Regulation Sandbox Program allows a US company, individual, or OSTP director to apply for a “waiver or amendment” of one or more federal agency regulations to “test, experiment, or temporarily provide” an AI product, AI service, or method of development. Your application must include information in various categories, including:
Contact information and business information, description of AI products, services, or development methods, specific regulations that the applicant is about to exempt or amend, and why such waiver or amendment is necessary, consumer benefits, business operational efficiency, economic opportunities, employment, and future AI products, services, services, or development methods, such as for salary for salary, to promote claims for future salary. The agency that has jurisdiction over the required period for mitigation, exemption or amendment, and the AI ​​products, services, or development methods.
Agent review and approval. The bill requires OSTP to submit an application within 14 days to a federal agency with jurisdiction over AI products, services, or development methods. In reviewing the AI ​​Sandbox Program Application, federal agencies should seek opinions from private sector and technical experts on whether the applicant’s plans benefit consumers, businesses, economics, or AI innovations, and whether potential benefits outweigh health and safety, economics, or consumer risks. The agency must approve or reject the application within 90 days, recording reasonably foreseeable risks, mitigation and consumer protection justifying agency approval, or reasons for the agency’s rejection. Applicants who are denied will be permitted to appeal to OSTP for reconsideration. An approved exemption or amendment is required by the applicant and, if approved by the OSTP, granted over a two-year term of office for up to four years.
Participant conditions and requirements. Participants with approved exemptions or amendments will be exempt from federal criminal, civil or agency enforcement of exemptions or amended regulations, but are subject to private consumer rights. Additionally, participants must report to OSTP and federal agencies any harm to health and safety, economic damage, or unfair or deceptive trade practices within 72 hours of the incident occurring, and make various disclosures to consumers. Participants must also submit repeat reports to OSTP during the exemption or amendment period. This should include the number of affected consumers, the risks and potential mitigation, unexpected risks that arise during deployment, adverse cases, and the benefits of exemptions or corrections.
Congress review. Finally, the Sandbox Act requires the OSTP director to submit regulations the director recommends to Congress for amendments, or to repeal “as a consequence of being able to operate safely” without those regulations under the Sandbox Program. The bill establishes expedited track procedures for joint resolutions approving such recommendations. This will immediately abolish regulations, if enacted, or adopt modifications recommended by OSTP.
Sandbox Act’s Regulatory Sandbox Program will be sunset in 12 years unless updated. The introduction of the Sandbox Act is based on the state pursuing its own AI-regulated sandbox program, including the Sandbox Program established under the Texas Liability AI Governance Act (“Traiga”), which was enacted in June, and the “AI Learning Institute Program” established under Utah’s 2024 AI Policy Act. In addition to accepting a coordinating reviews and accepting “co-applications” for participants in AI projects that benefit from “both federal and state regulatory relief,” the Sandbox Act requires that information be shared from these State AI Sandbox Programs if it is “similar or equivalent” to the Sandbox Act.
