USCO Standard

The Generative AI Copyright Disclosure Act, Explained

By Nora Whitfield · August 31, 2026

The Generative AI Copyright Disclosure Act, Explained

If you've searched for this bill recently, you've probably run into conflicting information, some of it out of date. Here's where things actually stand.

The original bill: introduced, and it died

Representative Adam Schiff of California introduced the Generative AI Copyright Disclosure Act, H.R. 7913, on April 9, 2024, during the 118th Congress. The bill was straightforward in its ambition: anyone who creates or materially alters a training dataset for a generative AI system would have to file a notice with the US Copyright Office, including a URL for the dataset and a sufficiently detailed summary of any copyrighted works used to build or update it. The filing deadline was set at 30 days before the model became available to the public, and the requirement was written to apply retroactively to models already on the market. Penalties for noncompliance started at $5,000 per violation, with no cap specified.

It drew real support from creative industry groups, songwriters and composers among them, who'd been asking for exactly this kind of transparency requirement. It also drew real pushback from AI developers and their advocates, who argued that comprehensively cataloging every copyrighted work in a modern training dataset, some of which run into the billions of documents and images, was practically unworkable at the scale the bill demanded.

The bill never got a floor vote. When the 118th Congress ended, H.R. 7913 died with it, the way any bill does that hasn't passed by the end of its congressional session. It has to be reintroduced from scratch in a new Congress to have any further life, and as of this writing, it hasn't been reintroduced under its original name and number.

What replaced it: the CLEAR Act

The idea didn't disappear. On February 10, 2026, Senators Adam Schiff and John Curtis introduced the Copyright Labeling and Ethical AI Reporting Act, known as the CLEAR Act, in the 119th Congress. It picks up the same core concept as the original disclosure bill: a notice requirement to the Register of Copyrights covering copyrighted works used to build generative AI models, aimed at giving rights holders visibility into whether their work was used for training.

The bipartisan pairing here is notable. Schiff is a Democrat from California. Curtis is a Republican from Utah. That's not the kind of pairing you get on a bill nobody in Congress cares about, and it suggests this general approach, transparency requirements for AI training data, has more durable political momentum behind it than a single bill's failure would suggest. Whether the CLEAR Act clears the same practical objections that sank the original disclosure bill, particularly the scale argument from AI developers, remains to be seen.

Why this keeps coming back

The underlying tension driving this legislation isn't going away regardless of which specific bill is on the table at a given moment. Rights holders, authors, musicians, visual artists, and their trade organizations, want to know whether and how their copyrighted work was used to train the AI models that are now competing with them commercially. AI developers argue that granular disclosure at the scale lawmakers are asking for is either technically infeasible or would expose competitively sensitive information about their training pipelines.

That fight is playing out in the courts too, not just Congress. The training-data lawsuits against OpenAI, Anthropic, Stability AI, and others are effectively asking judges to answer, case by case, questions that a disclosure law would make easier to litigate by simply requiring the information up front. You can see the shape of that litigation in our complete guide to every AI copyright lawsuit worth knowing, including the $1.5 billion Anthropic settlement, which turned in part on exactly the kind of provenance question a disclosure requirement would make far easier to establish.

What this means if you're a working creator right now

Nothing here changes your rights today. No disclosure law has passed yet, in either its original form or the CLEAR Act successor, so there's currently no legal requirement forcing AI companies to tell you whether your work was used in their training data. If you're trying to find out whether your material appears in a particular training set, that remains a matter of discovery in active litigation or voluntary disclosure by the company in question, not something you can compel through a filing requirement.

What is worth tracking is the direction this keeps moving. Two separate Congresses, two different bill numbers, the same core demand from the creative industries, and now a bipartisan sponsor pairing on the latest version. That's a pattern, not a one-off. If some version of a disclosure requirement eventually becomes law, it'll reshape how training-data infringement claims get litigated by giving plaintiffs a paper trail instead of forcing them to prove provenance through expensive discovery. Until then, the practical reality for AI companies and creators alike is the same as it's been throughout this fight: this stuff gets settled case by case, in court, on facts specific to each dispute.