LegisTrack
Back to all bills
S 2750119th CongressIntroduced

SANDBOX Act

Introduced: Sep 10, 2025
Technology & Innovation
Standard Summary
Comprehensive overview in 1-2 paragraphs

The SANDBOX Act would create a federal artificial intelligence regulatory sandbox program under the Director of the Office of Science and Technology Policy (OSTP). The idea is to let AI developers test AI products, services, or development methods on a limited basis by temporarily waiving or modifying certain existing regulatory provisions—so long as the testing is done in a controlled setting with appropriate safety and consumer protections. The program is designed to accelerate AI innovation and economic opportunity while giving regulators a structured way to study risks and determine what provisions might be safely relaxed or need adjustments in the future. The bill sets up a formal, multi-agency review process, requires public disclosure and ongoing reporting, and includes mechanisms for oversight, renewal, revocation, and judicial review. It also provides for a Congressional review process to consider repealing or amending covered provisions and includes a sunset after 12 years. Key elements include defined terms (what counts as an AI system, a covered provision, health/safety risk, etc.), a defined application process for waivers, agency review with input from industry and experts, binding written agreements, public disclosures to consumers, regular reporting to Congress, and a pathway for revoking waivers if terms aren’t met. The act anticipates coordination with state AI programs and limits on disclosing proprietary information. It ultimately seeks to balance innovation with consumer protection, while granting Congress a direct role in potentially amending or repealing waived provisions.

Key Points

  • 1Establishment of an Artificial Intelligence Regulatory Sandbox Program administered by the Director of the Office of Science and Technology Policy, offering temporary waivers or modifications of certain regulatory provisions to test AI products, services, or development methods on a limited basis.
  • 2Application, review, and decision process: applicants submit a detailed waiver/modification request outlining benefits, risks, mitigation plans, timelines, and applicable agencies; applicable agencies review with input from the private sector and technical experts, and must issue a written decision with described risks and mitigations; a Director-approved, written agreement is required before waivers become effective.
  • 3Oversight, enforcement, and renewal: waivers last for an initial 2-year period with possible renewals in up to four additional 2-year periods; the Director can revoke waivers for noncompliance; there are procedural requirements for notices, incident reporting (within 72 hours of harm or misconduct), and consumer disclosures; annual and periodic reporting to Congress on program activity, benefits, and harms.
  • 4Public transparency and consumer protection: participating entities must publicly disclose certain information to consumers (identity, program participation, testing status, non-immune status for non-waived provisions, and testing caveats); while protected from enforcement on waived provisions during the active period, they remain liable for other non-waived provisions or criminal liability not covered by the waiver.
  • 5Congressional review and sunset: Congress can consider amending or repealing waived provisions through a special joint-review process and joint resolutions of approval; the program terminates 12 years after establishment, ensuring a defined sunset and a built-in reevaluation point.

Impact Areas

Primary group/area affected: AI developers, AI product and service providers, and organizations testing AI development methods. They gain a structured pathway to test and refine AI with regulatory relief, under defined safety and consumer-protection conditions.Secondary group/area affected: Regulatory agencies responsible for enforcing the waived provisions, and the broader business community and investors involved in AI innovation. Agencies gain a framework for outside input and risk assessment; businesses may experience faster testing cycles but must comply with written agreements and reporting.Additional impacts: Consumers who will be exposed to AI products during sandbox testing, and state AI programs that may coordinate with federal testing efforts. There is potential for harmonization with state programs, clearer public information about ongoing tests, and a formal mechanism for Congress to influence or roll back regulatory requirements based on real-world testing outcomes.
Generated by gpt-5-nano on Oct 2, 2025