With generative AI on the rise, there have been growing concerns amongst copyright owners around the use of their materials in training such systems.
AI systems will commonly use banks of content to “learn” how to best operate and perform requested tasks. Such content may contain copyrighted works, such as sound recordings, books, articles, news reporting and more, and the owners of such works may not even know that they are being used as training materials. While developers often claim that their models are trained off of information in the public domain, the sheer expanse of data these systems access make it practically impossible to determine if copyrighted content is thrown into the mix.
The Generative AI Copyright Disclosure Act of 2024, brought by Rep. Adam Schiff of California, seeks to address some of these concerns by requiring persons creating a training data set, or altering an existing data set, to be used to build a generative AI system, to provide notice of their use of copyrighted works to the Register of Copyrights.
Here are some key take-aways from the proposed legislation:
- The required notice would provide a sufficiently detailed summary of the copyright works to be used as well as a URL for the data set if such is publicly available on the internet.
- The bill would put these disclosure requirements on new generate AI systems and would apply retroactively to systems already on the market.
- Those who fail to comply with the notice requirements would be subject to a civil penalty in an amount not less than $5,000.
As stated in a press release announcing introduction of the bill, Rep. Schiff believes that the proposed legislation “champions innovation while safeguarding the rights and contributions of creators” while “respecting creativity in the age of AI and marrying technical progress with fairness.”
The bill has received support from large organizations within the entertainment industry, including the Recording Industry Association of America and the Director’s Guild of America, among others.
In the absence of enacted legislation, many copyright owners have turned to the courts in search of protection and greater guidance. As it sits today, these cases are pending, meaning there are no clear answers to the many open questions presented by AI usage of copyrighted materials.
