The “Generative AI Copyright Disclosure Act of 2024” is a legislative proposal introduced by Representative Adam Schiff aimed at enhancing transparency in the AI industry. This act requires developers of AI models to disclose any copyrighted materials used in their training datasets to the Register of Copyrights, with detailed notifications including summaries of the “copyrighted works used.”
The necessity of this act was humorously underscored by an episode of the cartoon “South Park,” which depicted characters using a “ChatBot” to write essays and texts—a nod to the pervasive role of AI like ChatGPT in everyday tasks. Despite the satirical tone, the episode highlighted the reality of generative AI’s integration into daily life and the emerging issues related to the use of copyrighted content in AI training.
Overall, the proposed legislation targets the increasing issue of companies utilizing copyrighted materials—including web pages, books, and news articles—without authorization for training generative AI systems. The bill seeks to establish a process for transparent disclosure whenever copyrighted works are employed in the training of these systems. Its ultimate objective is to safeguard the rights of copyright owners.
Key Provisions of the Act
The bill mandates creators of AI training datasets to submit detailed notifications to the “Register of Copyrights.” This includes a comprehensive summary of any copyrighted works used in these datasets, like the URL of the website where the dataset was obtained. Notices must be submitted at least 30 days before the AI system is launched commercially or within 30 days post-enactment if the system was already in use.
Implications for Compliance and Enforcement
Failure to comply with these requirements can result in civil penalties “in an amount not less than $5,000.” The Act also stipulates the establishment of a public online database to maintain transparency, listing every notice filed. This database will be crucial for copyright holders to monitor the use of their works and for the public to understand the content feeding into the generative AI system used.
Defining the Landscape
The Act defines key terms essential to its application, such as “Artificial Intelligence,” “Generative AI Model,” “Generative AI System,” and “Training Dataset.” These definitions are crafted to encapsulate a broad range of AI functionalities that involve human-like cognitive processes, emphasizing the Act’s comprehensive approach to generative AI.
Set to take effect 180 days post-enactment, the “Generative AI Copyright Disclosure Act of 2024” represents a stride towards embedding accountability to generative AI development. By fostering a regulated environment, this legislation seeks to protect copyright owners while bolstering the ethical development of AI technologies. While not comprehensive enough to tackle all issues surrounding the growing use of AI systems, the bill does attempt to be a starting point for ensuring that innovations in AI are aligned with existing intellectual property (IP) laws.
Takeaways
In view of the potential impact the “Generative AI Copyright Disclosure Act” may have if enacted, the following are items organizations may want to consider implementing:
- Review Your AI Training Data: Examine any copyrighted works used in your AI models’ training datasets. During the review, companies should attempt to document what copyrighted material is being used and attempt to assess the copyright owners and any licensing status for the work.
- Consent of Copyright Owners: Take the initiative to contact the owners of the copyrighted works used in your training data. Try and secure the appropriate permissions or licenses for their use and show a commitment to working collaboratively with creators and compensating them appropriately.
- Training Data Protocol: Set up a detailed process to monitor, track, and report the use of copyrighted works in your AI training data. This should include mechanisms for submitting the necessary disclosures to the Copyright Office. This may also include the use of 3rd party tools or software that can assist in identifying copyrighted works and the potential owners. Accurately document and publicly share the required details under the Act, including a summary of the copyrighted works utilized, while avoiding any practices that could mislead about the use of protected content.
- Ongoing Education: Engage with a knowledgeable intellectual property lawyer to get continuous advice and assistance in navigating the complexities of intellectual property and generative AI laws which are still in the proposed state, like the one discussed above, or which have already been signed into law (e.g., Utah’s Generative AI law).
*****
For assistance in understanding the immediate impact of IP or generative AI laws, please contact John Rondini, co-chair of Brook Kushman’s Cybersecurity and Data Privacy Group, at 248.226.2913 or at jrondini@brookskushman.com.
