Issues, policy, and open areas of law
There can be no question that generative AI is transforming the landscape of copyright law. From whether the training ingestion of copyrighted content is permissible fair use, to the arguably derivative nature of generative-AI’s output, to nuanced – and internationally variable – registrability requirements for works created utilizing generative AI, the AI-related issues at the forefront of copyright law are diverse and unse led. What should companies bear in mind as they navigate this evolving landscape? And how can they best leverage the benefits of this technology while mitigating associated copyright risk?
One headline issue surrounds whether the ingestion of copyrighted material, absent permission, for the purpose of training generative AI models is an impermissible use under copyright law that infringes the copyrights held in the underlying material. In the U.S., several lawsuits including class actions have been filed on these grounds, with fair use defenses being asserted in response. These issues are live before the courts and yet to be decided. In the EU, legislation has already created a statutory text and datamining exception which allows the use of copyrighted material for the purpose of training generative AI in certain cases, unless the copyright owner has reserved this right. However, it is yet to be clarified how this applies in practice and how copyright owners can opt out of their material being used. Additionally, the EU AI Act has now entered into force, and imposes transparency obligations on AI providers to inform about the AI-training. The constantly new developments in the U.S. and in Europe should closely be followed by companies developing and deploying generative AI.
A second issue concerns whether, because of the way generative AI models are trained, their output is an infringing derivative of the copyrighted works they ingest. While any risk will be circumstance and jurisdiction specific, including with respect to the particular content ingested, commands used, and output generated, both generative AI developers and users should be aware of the potential liabilities – including direct, contributory, and vicarious – associated with arguably infringing generative AI output.
Third, copyright registrants and litigants are well advised to stay abreast of the nuanced and evolving registrability requirements for works involving generative AI, which are likely to implicate copyright litigation more broadly. In March 2023, the U.S. Copyright Office issued a statement of policy applying a human authorship requirement for copyrightability, concluding that works created by generative AI in response to human prompting do not meet this requirement. That policy statement also requires that AI- generated material be disclosed when applying for copyright registration, that previously filed applications which do not disclose the use of AI be corrected, and that supplemental registrations identifying and disclaiming any AI generated material contained within previously registered works be filed. Since that time, the Copyright Office has commenced its roll-out of a series of reports relating to the implications of AI on copyright law and policy, and which is widely anticipated to update these requirements. Pending the nature of those updates, and combined with varying registrability policies in other Berne Convention jurisdictions, such policy and requirements may give rise to significant avenues for discovery and attacks on validity in copyright litigation moving forward.