TRAIN Act
Summary
What This Bill Does
The TRAIN Act adds a new title 17 section 514 on subpoenas for artificial intelligence training materials. It defines AI, AI model, generative AI model, developer, substantial modification, and training material. A legal or beneficial owner of an exclusive copyright right, or someone authorized to act for the owner, may request that a U.S. district court clerk issue a subpoena to a developer for copies of, or records sufficient to identify with certainty, the copyrighted works or portions likely owned or controlled by that owner and used to train a generative AI model. The requester must have a subjective good-faith belief that the developer used one or more of the owner's works for training. The request is limited to the requester's own works and requires a proposed subpoena plus a sworn declaration stating the belief, purpose, and rights-protection use. If the papers are proper, the clerk must issue and sign the subpoena. Developers must expeditiously disclose the requested copies or records. Recipients of records have a confidentiality duty. Federal Rule of Civil Procedure subpoena procedures and remedies apply where not displaced. Failure to comply creates a rebuttable presumption that the developer made copies of the copyrighted work. Courts may impose Rule 11-style sanctions for bad-faith subpoena requests.
Who Benefits and How
Copyright owners, authors, artists, publishers, photographers, musicians, media companies, and authorized rights representatives benefit from a court-clerk subpoena path to identify whether their works were used in generative AI training. Federal courts and copyright litigants gain a structured evidence mechanism before or alongside infringement disputes. Developers who receive bad-faith subpoenas benefit from a sanctions remedy and the limitation to the requester's own works.
Who Bears the Burden and How
Generative AI developers, AI model owners, training dataset curators, State or local government AI developers, district court clerks, federal courts, and subpoena recipients face new disclosure, record-search, confidentiality, compliance, and litigation burdens. Developers that fail to comply face a rebuttable presumption that they copied the copyrighted work. Copyright owners must file sworn declarations, keep disclosed records confidential, and risk sanctions for bad-faith requests.
Key Provisions
- Creates a Copyright Act subpoena process for identifying copyrighted works used to train generative AI models.
- Limits subpoenas to works likely owned or controlled by the requesting copyright owner.
- Requires a proposed subpoena and sworn declaration of subjective good-faith belief and rights-protection purpose.
- Requires developers to expeditiously disclose copies or records sufficient to identify the training materials.
- Imposes confidentiality duties on copyright owners receiving records.
- Creates a rebuttable presumption for developer noncompliance and sanctions for bad-faith subpoena requests.
Evidence Chain:
This summary is generated from the full bill text using AI analysis. Expand "Detailed Analysis" below for identified beneficiaries/burden bearers with clause-level evidence links.
At a Glance
What This Bill Does
Creates a new Copyright Act subpoena process allowing a copyright owner or authorized representative with a subjective good-faith belief that a developer used the owner's works to train a generative AI model to obtain from a federal district court clerk a subpoena for copies or records sufficient to identify the owner's training materials, while limiting requests to the requester's own works, imposing confidentiality duties, applying subpoena enforcement rules, creating a rebuttable presumption when developers fail to comply, and allowing sanctions for bad-faith requests.
Key Policy Areas
Technology, Intellectual Property, Courts
Primary Purpose
Creates a new Copyright Act subpoena process allowing a copyright owner or authorized representative with a subjective good-faith belief that a developer used the owner's works to train a generative AI model to obtain from a federal district court clerk a subpoena for copies or records sufficient to identify the owner's training materials, while limiting requests to the requester's own works, imposing confidentiality duties, applying subpoena enforcement rules, creating a rebuttable presumption when developers fail to comply, and allowing sanctions for bad-faith requests.
Policy Domains
Substantive provisions
Identified Gains
- Copyright owners
- Authors
- Artists
- Publishers
- Photographers
- Musicians
- Authorized rights representatives
Identified Costs
- Generative AI developers
- AI model owners
- Training dataset curators
- District court clerks
- Federal courts
- Copyright owners filing subpoenas
Sponsors
Legislative Progress
In CommitteeMs. Dean of Pennsylvania (for herself and Mr. Moran) introduced …
Referred to the House Committee on the Judiciary.
Introduced in House
Stakeholder Effects
cui bono?How this legislation distributes effects. Mention counts reflect frequency, not effect magnitude.
Artists, Authors, Copyright owners
Positive-direction: Artists, Authors, Copyright owners, Musicians, Photographers, Publishers
Negative-direction: Copyright owners filing subpoenas
AI model owners, Generative AI developers, Training dataset curators
Bill Structure & Actor Mappings
Who is "The Secretary" in each section?
We use a combination of our own taxonomy and classification in addition to large language models to assess meaning and potential beneficiaries. High confidence means strong textual evidence. Always verify with the original bill text.
Learn more about our methodology