The bill strengthens copyright owners' ability to probe AI training data through expedited subpoenas and narrower scope rules, but imposes disclosure costs and confidentiality risks on AI developers that could harm competition and chill innovation.
Copyright owners (including small-business owners and nonprofits) can compel AI developers to disclose records about whether and how their works were used to train generative models.
Provides a faster, court-administered mechanism (clerk-issued subpoenas) to obtain disclosures from AI developers, reducing delay and lowering enforcement friction for rights-holders.
Limits subpoenas to works likely owned or controlled by the requester, reducing the risk of broad fishing expeditions into unrelated copyrighted materials.
AI developers (including small firms and tech workers) may incur significant compliance costs and administrative burdens to gather and produce training records on short timelines.
Forcing disclosure of training materials risks revealing trade secrets or proprietary model details, creating competitive harm for developers and undermining commercial incentives.
The provision excludes noncommercial end users from developer protections and may prompt developers to restrict access or collaboration, chilling innovation, research, and nonprofit use.
Based on analysis of 2 sections of legislative text.
Inserts a new Title 17 provision authorizing subpoenas for records related to AI models and sets out broad definitions for AI, models, developers, and training material.
Introduced January 22, 2026 by Madeleine Dean · Last progress January 22, 2026
Adds a new provision to federal copyright law creating a subpoena authority for "copies or records relating to artificial intelligence models" and supplies detailed definitions for terms like "artificial intelligence," "artificial intelligence model," "generative artificial intelligence model," "developer," "substantially modify," and "training material." The text provided includes the definitions and the structural insertion into Title 17 but does not include the operative subpoena language, specific requirements, deadlines, agency roles, or funding details. Because the bill includes only the structural addition and definitions as provided, key practical effects—such as who may issue subpoenas, what exactly must be produced, how trade secrets or privacy are handled, and enforcement procedures—cannot be determined from the supplied text. The definitions, however, are broad and would cover a wide range of models, private-sector developers, and some government actors if the missing subpoena provisions allow compulsion of records.