TRAIN Act
- Bill Number
- S. 2455
- Origin Chamber
- Senate
- Congress
- 119th Congress, Session 1
- Policy Area
- Commerce
- Status
- Introduced
- Latest Action
- 2025-07-24: Read twice and referred to the Committee on the Judiciary.
- Last Updated
- 2026-01-26T15:08:13Z
AI-Generated Summary
Purpose
The legislation establishes a new administrative subpoena process under copyright law to enable copyright owners to obtain information about whether their works were used to train generative artificial intelligence models.
Key Provisions
- Definitions: The bill defines key terms, including artificial intelligence, artificial intelligence model, developer (covering persons or state/local government agencies that design or train such models, but excluding noncommercial end users), generative artificial intelligence model, substantially modify, and training material.
- Subpoena Request Process: Copyright owners (or their authorized representatives) may file a request with a U.S. district court clerk for a subpoena to an AI developer. The request requires a sworn declaration of a subjective good faith belief that the owner's works were used in training, along with a proposed subpoena.
- Scope Limitations: Subpoenas are restricted to the requester's own copyrighted works and aim to obtain copies or identifying records of training materials.
- Developer Obligations: Upon receipt, developers must promptly disclose the requested records. Noncompliance creates a rebuttable presumption that the developer copied the copyrighted work.
- Confidentiality and Enforcement: Recipients of records must maintain confidentiality. Procedures follow the Federal Rules of Civil Procedure for subpoenas, with sanctions available for bad faith requests under Rule 11.
- Effective Date: The new section takes effect upon enactment.
Significant Changes to Existing Law
This bill adds a new Section 514 to Chapter 5 of Title 17 of the U.S. Code, creating a specialized subpoena mechanism for AI training disclosures. It is not present in current copyright law and provides a streamlined, court-clerk-issued process outside traditional litigation, while incorporating a rebuttable presumption of infringement for noncompliance.
Potential Impacts
- On Government Agencies: State and local agencies that qualify as developers may face disclosure requirements when subpoenaed.
- On Citizens: Copyright owners gain a tool to investigate potential unauthorized use of their works in AI training, potentially aiding enforcement of rights without full lawsuits.
- On International Relations: No direct provisions address foreign entities, but the process could apply to U.S.-based developers handling international training data, indirectly affecting global AI practices.
- Broader Effects: AI model developers may incur compliance costs and face increased scrutiny, while encouraging transparency in training datasets.
Main Stakeholders Affected
- Copyright owners and their representatives.
- AI developers, including companies and state/local government agencies involved in model design or training.
- U.S. district courts responsible for issuing subpoenas.
- Noncommercial end users of AI models (explicitly excluded from developer status).
Notable Legal, Constitutional, or Political Implications
The bill introduces a rebuttable presumption of copying upon subpoena noncompliance, which could shift burdens in future litigation. It relies on existing Federal Rules of Civil Procedure for enforcement, potentially raising questions about due process in administrative-like subpoenas. The good faith requirement and sanctions for misuse aim to prevent abuse, but the process may lead to increased court involvement in copyright-AI disputes.
This summary was generated by AI and may contain inaccuracies. Refer to the official source document for the authoritative text.
Sponsor
Cosponsors (3)
Sen. Blackburn, Marsha [R-TN], Sen. Hawley, Josh [R-MO], Sen. Schiff, Adam B. [D-CA]
Recent Actions
- 2025-07-24: Read twice and referred to the Committee on the Judiciary.
- 2025-07-24: Introduced in Senate
Bill Versions
- Transparency and Responsibility for Artificial Intelligence Networks Act — issued 2025-07-24 — PDF (8 pages)