When it comes to law and technology, the law often lags far behind. In the case of generative AI systems, it appears that at least one legislator is attempting to close the gap. A new bill introduced by Representative Adam Schiff (CA) would require companies to disclose any copyrighted materials used to train their generative AI systems.
The Generative AI Copyright Disclosure Act of 2024 requires that a person who creates or alters a training dataset used in building a generative AI system disclose to the copyright Register a notice of copyrighted works used. The notice must contain a detailed summary of any copyrighted works used in the training or altering of a dataset and the URL for such dataset. The notices would then be maintained in a publicly accessible database. Companies must submit the notice not later than 30 days before releasing the generative AI system in question to the public. The bill’s requirements would also apply retroactively to previously released generative AI systems. The penalty for failing to comply is an amount not less than $5,000.
This Act seeks to address the tension between generative AI systems and copyright owners. Generative AI systems are trained on bodies of information known as datasets, which allow the system to function. However, the information contained within the datasets often tends to include copyrighted work that was used without permission. In light of this, companies behind generative AI systems are facing a slew of legal challenges.
The bill has generated support from the Recording Industry Association of America, the Director’s Guild of America, Professional Photographers of America, SAG-AFTRA, Writers Guild of America West, and Writers Guild of America East, among others.