Copyright owners get easy subpoena power over AI training data—with minimal oversight.
S. 2455 — TRAIN Act · Filed by Peter Welch (D-VT) · 3 cosponsors · Introduced Jul 24, 2025 · Referred to committee
Your members of Congress
Enter a ZIP to see where your representative and both senators stood on this bill.
Looked up on this device — your ZIP is never stored on our servers.
What it does
This bill creates a legal mechanism allowing copyright owners to subpoena AI developers and demand disclosure of which copyrighted works were used to train their AI models. A copyright owner files a sworn declaration with a court clerk stating they have a good-faith belief their work was used in training, and the clerk must issue the subpoena automatically if the paperwork is in order. The developer must then disclose copies or records identifying the training material, with confidentiality protections and sanctions available for bad-faith requests.
Why we flagged it
The bill's core function is to establish an administrative subpoena process specifically for copyright owners to compel disclosure of training data used in AI models. It is not a ban, licensing scheme, or liability rule—it is a discovery tool with a very low threshold for issuance.
What the text implies
- The 'subjective good faith belief' standard requires no evidence or probable cause before a clerk must issue a subpoena—developers face mandatory disclosure obligations triggered by unverified assertions, creating asymmetric discovery risk.
- Developers must disclose 'all records' identifying training material, potentially exposing proprietary training datasets, model architecture details, and competitive sourcing strategies to copyright claimants who may be competitors.
The full analysis lists 5 implications of this text.
Who stands to gain
copyright holders and licensing entities; entertainment and publishing companies; litigation support and discovery service providers