Generative AI Copyright Disclosure Act
The Generative AI Copyright Disclosure Act of 2024 was a U.S. federal bill, H.R. 7913, introduced in the House of Representatives on April 9, 2024. As introduced, it proposed disclosure obligations relating to copyrighted works used in training datasets for generative AI systems.
The bill is useful to understand because it reflects a broader concern: when generative AI systems are built or used, what should be disclosed, by whom, and at what level of detail?
What it is
The Generative AI Copyright Disclosure Act was a proposed federal transparency measure. In the version introduced in Congress, it focused on disclosure related to copyrighted works included in training datasets used for generative AI systems.
So the term refers to a specific proposed bill about training-data disclosure, not to a generic industry standard and not to a settled rule that automatically applies to every organization using AI.
What it is not
It is not a general answer to all copyright questions around AI-generated content.
It is not the same thing as an AI use statement, an AI disclosure statement, or a provenance disclosure for a specific output.
And it is not a complete framework for explaining how a particular document, design, report, or software deliverable was created.
Status matters
Congress.gov lists H.R. 7913 as introduced on April 9, 2024, referred to the House Judiciary Committee, and not enacted. That means it should be understood as a proposed legislative approach, not as a currently effective nationwide requirement.
What the bill proposed
As introduced, the bill would have required a notice to be submitted to the Register of Copyrights for training datasets used in building generative AI systems.
The bill text said that the notice should include a sufficiently detailed summary of copyrighted works used in the relevant training dataset, and the URL for the dataset if it was publicly available on the internet.
The timing provisions in the introduced text also contemplated notices before consumer release of a new system, and post-effective-date notice requirements for systems already made available before the act took effect.
Why it matters
The bill matters because it highlights a real distinction in AI transparency discussions. One kind of disclosure concerns model development and training data. Another concerns the origin of a specific output released to customers, publishers, reviewers, or counterparties.
Those are related issues, but they are not the same issue.
Questions organizations still need to answer
Even if a training-data disclosure rule existed, organizations would still need to answer broader questions such as:
- Was AI used in producing this work?
- What role did AI systems play in the workflow?
- Who reviewed and approved the final result?
- What record exists to explain the creation process later?
Those are output-level questions about process and responsibility, not just questions about model training.
Training-data disclosure is different from output disclosure
The bill focused on disclosure related to training datasets used in building generative AI systems. That is different from the kind of disclosure many organizations need when they release a document, report, design, or other work product created with AI assistance.
In those cases, the central questions are often:
- Was AI used in creating this work?
- What role did AI play in the workflow?
- Who reviewed and approved the final result?
Those are provenance and process questions, not just training-data questions.
How this relates to Provenance Disclosure
Provenance Disclosure is designed to generate a structured record describing how a specific work was created. That is a different function from a training-data disclosure rule aimed at model developers.
If you need to document the creation of a particular document, report, design, software deliverable, or other work product, you are usually dealing with an output-level provenance problem, not a training-dataset notice problem.
In practice, an organization might care about both issues at once: model-training transparency at the system level, and output-level provenance for the specific artifacts it releases.
When Provenance Disclosure is relevant
If your question is not “what did a model developer use to train this system?” but instead “how was this specific work created and reviewed?”, a provenance disclosure is usually the more relevant format.
That is especially true for publishing, procurement, internal governance, and buyer review contexts where the focus is on the origin, review, and approval of a particular output.
Generate a structured disclosure
If you need a formal record describing how AI tools were involved in creating a specific document, product artifact, or creative work, generate a structured provenance disclosure.