USA Trends and Developments Contributed by: Anne Shea Gaza, Adam W Poff, Pilar G Kraman and Robert M Vrana, Young Conaway Stargatt & Taylor, LLP
The Impact of Generative AI on Trade Marks and Copyrights The rapid evolution of generative artificial intel - ligence is beginning to affect many industries, and intellectual property is no exception. Gen - erative AI, which refers to AI systems capable of creating new content, has far-reaching impli - cations for how creators and businesses can protect their trade marks and copyrights. As generative AI becomes increasingly prevalent, it also challenges assumptions about author - ship and ownership of intellectual property – for example, can an AI model be an “author” for purposes of copyright law? In the short-term, however, legal developments are more likely to be focused on the use by generative AI models of trade marks and copyrighted works owned by others. Moreover, recent decisions by the US Supreme Court addressing “fair use” in both US copyright law and US trade mark law have created an environment ripe for litigation over generative AI models. How will traditional legal frameworks apply to, and adapt in light of, these new technologies? Understanding generative AI Generative AI involves machine learning models that can generate content such as text, imag - es, audio, and even videos. These systems are trained on large sets of data, which allows them to learn the styles, patterns, and structures of existing content and produce output based on what they have learned. The produced output is therefore new, unique, and produced without direct human input, even though the data on which the generative AI model was trained was pre-existing content created by humans. For example, if a user asks a generative AI model to produce a poem in the form of a haiku, the model will be able to do so because it has pre - viously “learned” from ingesting thousands of haikus. Common examples include ChatGPT, a
large language model that generates text, and DALL E, a text to image model that generates images from text prompts. Both of these exam - ples were developed by OpenAI. Copyright law: infringement and derivative works The US Copyright Act is designed to protect the rights of creators in their original works of authorship, granting them exclusive rights to reproduce, distribute, display, and perform their works. A copyrighted work, thus, cannot be used by others without the permission of the work’s owner. For generative AI, the question arises whether the use of copyrighted works to “train” the AI model constitutes copyright infringement because the works used for training are copied and used, usually without permission. In other words, does the “learning” process of a generative AI model constitute a violation of copyrights? Similarly, only copyright owners have the right to make “derivative works” based on their cop - yrighted works. The output of a generative AI model, which is inherently based on other works that were used to train the model, can be seen as inherently at least somewhat derivative. But is it based closely enough on any one copyrighted work to be deemed a “derivative” of that copy - righted content? The output of generative AI models may closely resemble or reproduce ele - ments of certain copyrighted works ingested by the model during training. On the other hand, the output could also be sufficiently “transforma - tive” that it cannot be said to be derivative of any particular content. In other words, does the out - put itself of the generative AI model – long after the training process has occurred – constitute a copyright violation if it is too similar to an earlier copyrighted work?
751 CHAMBERS.COM
Powered by FlippingBook