Copyright Implications of Generative AI Use

Copyright uncertainty is among the main obstacles to the confident adoption of General-Purpose AI (GPAI) systems by companies. Cutting-edge GPAI systems are typically trained on text, images, and code available on the internet, some of which is under copyright. However, the legality of AI systems training on copyrighted information depends on whether such information falls within an exception to copyright protection, such as "fair use" in the US or "text and data mining" in the EU.

For companies using GPAI systems, the copyright considerations at play include:

• Human authorship requirement: In many jurisdictions, copyright protection is only available for materials authored by humans. So, companies need to interpret what constitutes meaningful human authorship.
• Similarity to source material: GPAI systems sometimes generate outputs that are similar to specific copyrighted materials within their training data, which could subject the user to potential infringement claims.
• Differing copyright interpretations by jurisdiction: Copyright-related statutes, regulatory guidance, and case law sometimes differ across jurisdictions, creating additional complexity for companies that operate internationally.

Companies can reduce the copyright risks of using GPAI systems by:

• Understanding provider indemnification: Many major providers of GPAI systems indemnify customers against copyright infringement claims related to system outputs. Such indemnification often varies by the version and tier of the AI system used, among other factors.
• Asking providers about training data: Companies should know where the providers of the GPAI systems they use sourced their training data and confirm that the data was legally acquired.
• Sharing the company position on meaningful human authorship: Anyone using GPAI systems within the company should be aware of the company’s position on what constitutes meaningful human authorship, which should be communicated in simple and clear language.
• Monitoring legal developments: Though court cases generally take time to resolve, they can have significant implications as soon as they are decided. Companies should monitor relevant cases, as well as statutory and regulatory developments, so that they can rapidly incorporate new information into their AI governance programs.

(Note that the copyright issues discussed here are separate from ensuring that a company’s proprietary, personal, or otherwise valuable information is not used to further train GPAI systems, which is beyond the scope of this article.)

For the purposes of the EU AI Act, which covers copyright considerations for GPAI systems in Articles 53(1)(c) and (d), the EU AI Office is drawing up a GPAI Code of Practice that will clarify copyright-related requirements for GPAI providers.

The second of three drafts of the GPAI Code of Practice, published on December 19, 2024, includes several copyright-related commitments, including that GPAI providers should adopt, and publicly summarize, an internal copyright policy that ensures compliance with EU law, prevents the generation of copyright-infringing content, and puts in place measures to identify and comply with rights reservations.

Given the significant changes between the first and second drafts of the GPAI Code of Practice, companies should wait until the third and final draft is due to be published at the end of April 2025 to better understand how GPAI providers are expected to address copyright concerns. 

Enzai is here to help

Enzai’s AI GRC platform can help your company deploy AI in accordance with best practices and emerging regulations, standards and frameworks, such as EU AI Act, the Colorado AI Act, the NIST AI RMF and ISO/IEC 42001. To learn more, get in touch here.