Currently, two bipartisan bills in Congress address the use of copyrighted works in generative artificial intelligence (AI) training. Both share a common goal of improving transparency in how AI models are trained on copyrighted material, but they take different regulatory approaches. At present, both bills remain in the early stages of the legislative process.

TRAIN Act

Representatives Nathaniel Moran (R-TX) and Madeleine Dean (D-PA) introduced the Transparency and Responsibility for Artificial Intelligence Networks (TRAIN) Act in the U.S. House of Representatives (H.R. 7209). The bill has been introduced and referred to committee but has not yet advanced to hearings, markup, or a floor vote. A similar version has also been introduced in the Senate. As of now, the legislation remains pending in committee.

The TRAIN Act is enforcement-focused. It would allow copyright owners to seek administrative subpoenas requiring AI developers to disclose whether and how copyrighted works were used in training AI systems. Developers would have a legal obligation to respond.

This mechanism is reactive. The creator must first suspect that their work has been used before seeking disclosure. The bill is designed to help creators investigate potential infringement and pursue litigation if necessary.

Practical effect if enacted
If passed, the TRAIN Act would create a formal legal pathway for creators to obtain training data information that is currently difficult or impossible to access. It would likely increase litigation or pre-litigation discovery related to AI training practices and shift some compliance burden to AI developers to maintain adequate training data records. However, it would not require routine public disclosure of training datasets.

CLEAR Act

Senators John Curtis (R-UT) and Adam Schiff (D-CA) introduced the Copyright Labeling and Ethical AI Reporting (CLEAR) Act in the U.S. Senate. Like the TRAIN Act, it has been introduced and referred to committee but has not advanced further in the legislative process.

The CLEAR Act is disclosure-focused. It would establish a mandatory reporting system requiring AI developers to disclose which copyrighted works are used to train generative AI models, typically by filing notices before releasing a model. This would create a structured transparency regime so creators and the public could see how training data is used.

Unlike the TRAIN Act, which operates after suspected use, the CLEAR Act provides proactive transparency prior to or at the time of model release.

Practical effect if enacted
If passed, the CLEAR Act would impose ongoing compliance obligations on AI developers, including pre-release reporting requirements. It would likely increase operational costs related to dataset documentation, tracking, and disclosure. For creators, it would provide earlier visibility into whether their work is included in training datasets, potentially enabling licensing negotiations or enforcement before disputes escalate.

Summary

Both bills aim to increase transparency in AI training practices involving copyrighted works. However, they differ in structure:

• The TRAIN Act creates a reactive enforcement pathway through an administrative subpoena system.
• The CLEAR Act establishes a proactive reporting framework requiring disclosure before model release.

Both remain in the early legislative stages, and neither has advanced beyond committee referral. If enacted, TRAIN would primarily strengthen investigative tools for creators, while CLEAR would fundamentally alter compliance obligations by establishing systemic transparency requirements for AI developers.

We will continue to monitor legislative developments related to these bills. As Congress refines its approach to AI and intellectual property, we will provide updates on significant changes and practical implications for creators and AI developers alike.