SANDBOX Act

Bill Number

S. 2750

Origin Chamber

Senate

Congress

119th Congress, Session 1

Policy Area

Science, Technology, Communications

Status

Introduced

Latest Action

2025-09-10: Read twice and referred to the Committee on Commerce, Science, and Transportation.

Last Updated

2025-09-29T14:54:37Z

AI-Generated Summary

Purpose of the Legislation

The SANDBOX Act aims to promote innovation in artificial intelligence (AI) by creating a temporary "regulatory sandbox" program. This allows businesses and the government to test AI products, services, or development methods with limited waivers from certain federal regulations, while ensuring safeguards against health, safety, economic, or deceptive practice risks. The goal is to encourage AI development, expand economic opportunities, create jobs, and position the United States as a leader in AI without immediately imposing full regulatory burdens.

Key Provisions

• Establishment of the Program: The Director of the Office of Science and Technology Policy (OSTP) must create the AI Regulatory Sandbox Program within one year of enactment. It enables applicants (businesses or the Director) to request temporary waivers or modifications of federal rules (called "covered provisions," which include regulations, guidance, and statutorily required rules) for testing AI on a limited basis.

• Application and Review Process:
• Applicants submit detailed forms describing the AI innovation, benefits (e.g., consumer advantages, efficiency, job creation), risks (health/safety, economic damage, unfair/deceptive practices), mitigation plans, and requested waiver duration.
• OSTP assists applicants and forwards applications to relevant federal agencies (e.g., those enforcing the rules) for review within 90 days (with possible 30-day extension).
• Agencies assess benefits versus risks, consult experts/private sector, and approve/deny via a record explaining decisions. If all agencies approve, OSTP grants the waiver; partial approvals are possible.
• Appeals go to OSTP, with decisions within 60 days.

• Waivers and Operations:
• Waivers last 2 years initially, renewable up to four more 2-year periods (total 10 years), unless risks change or terms are violated.
• Participants enter written agreements with OSTP and agencies outlining waived rules and risk mitigations; must report incidents (e.g., harm to consumers) within 72 hours.
• Revocation possible for non-compliance, with 30+ days to correct.
• Director can submit applications to advance AI; approved ones can be used by others via a standardized process.

• Consumer Protections and Disclosures:
• No waiver of existing consumer rights to sue for damages or seek remedies.
• Participants must publicly disclose program participation, risks, lack of immunity from non-waived rules, and contact info for complaints (e.g., to the National AI Initiative Office) via websites or similar means.
• No criminal/civil enforcement for waived rules during compliance, but full liability remains for other violations.

• Reporting and Oversight:
• Participants submit reports at 40 days, halfway, and end of periods on consumers affected, risks/mitigations, incidents, and benefits.
• OSTP reports annually to Congress on applications, approvals, public benefits/harms, waived rules, and all related materials.
• Coordinates with state AI sandbox programs for information sharing, joint applications, and harmonized testing.

• Congressional Review (Section 703): Annually, OSTP submits recommendations to Congress for repealing or amending rules based on sandbox experiences (e.g., if safe operation without them is proven). Congress can approve via expedited joint resolutions, with fast-track procedures in both chambers to bypass some delays.

• Other Details:
• Program sunsets (ends) 12 years after establishment.
• No requirement to disclose proprietary info; does not affect non-waived laws.
• Judicial review available for key decisions (e.g., denials, revocations) as final agency actions.

Significant Changes to Existing Law

• Amends the National Science and Technology Policy, Organization, and Priorities Act of 1976 by adding a new Title VII specifically for the AI sandbox.
• Introduces a novel federal mechanism for temporary regulatory relief tailored to AI, unlike existing sandboxes (e.g., in fintech) that are agency-specific and not centralized under OSTP.
• Adds congressional fast-track procedures for approving regulatory reforms based on sandbox data, streamlining changes to federal rules without standard legislative hurdles.
• Defines AI and related terms by reference to the 2020 National AI Initiative Act, ensuring consistency, and specifies risks (e.g., "health and safety risk" includes harm to human life or health, even before birth).

Potential Impacts

• On Government Agencies: Increases coordination among agencies (e.g., FTC for deceptive practices, others for sector-specific rules) via reviews and agreements; may lead to streamlined regulations if Congress acts on OSTP recommendations, reducing enforcement burdens long-term but requiring new advisory processes.

• On Citizens (Consumers): Enables faster access to innovative AI tools (e.g., in healthcare or services) with potential benefits like efficiency and jobs, but introduces controlled risks during testing; protections ensure transparency and retained legal rights, though participants must monitor for harms.

• On International Relations: Could enhance U.S. competitiveness in global AI development by accelerating innovation, potentially influencing international standards or trade in AI tech; no direct foreign policy provisions, but success might encourage similar programs abroad or bilateral coordination.

Main Stakeholders Affected

• AI Developers and Businesses: Primary beneficiaries, gaining regulatory flexibility to test innovations; must comply with reporting and disclosures.
• Federal Agencies: OSTP leads, but enforcement agencies (e.g., FTC, FDA if health-related) review applications and provide input, potentially shifting workloads.
• Consumers and the Public: Directly impacted by tested AI products/services; gain innovation opportunities but face disclosed risks.
• Congress and States: Congress gains tools for evidence-based regulatory reform; states with AI programs benefit from federal coordination.
• Technical Experts and Private Sector: Involved in agency consultations for reviews.

Notable Legal, Constitutional, or Political Implications

• Legal: Provides limited immunity from enforcement for waived rules during compliance, but preserves consumer lawsuits and non-waived liabilities; enables judicial review under the Administrative Procedure Act for key decisions, ensuring accountability. The rule of construction protects trade secrets, aligning with intellectual property laws.

• Constitutional: Fast-track congressional procedures exercise rulemaking powers of each chamber, potentially raising separation-of-powers questions if seen as delegating too much to OSTP, but include direct congressional approval for changes, maintaining legislative oversight.

• Political: Promotes pro-innovation policy by balancing deregulation with safety, potentially partisan (e.g., favoring tech industry growth); annual reports and sunset clause allow periodic evaluation, reducing long-term entrenchment. Could set precedent for sandboxes in other emerging tech areas, influencing broader regulatory debates.

This summary was generated by AI and may contain inaccuracies. Refer to the official source document for the authoritative text.

Sponsor

Sen. Cruz, Ted [R-TX]

Recent Actions

• 2025-09-10: Read twice and referred to the Committee on Commerce, Science, and Transportation.
• 2025-09-10: Introduced in Senate

Bill Versions

• Strengthening Artificial intelligence Normalization and Diffusion By Oversight and eXperimentation Act — issued 2025-09-10 — PDF (41 pages)