OpenAI, the driving force behind ChatGPT and its large language models (LLMs) GPT-3.5 and GPT-4, is currently facing two copyright lawsuits. These lawsuits allege that OpenAI used copyrighted materials in training data for its AI models without permission. The plaintiffs include a pair of U.S. authors and a group that includes renowned comedian and actor Sarah Silverman. OpenAI has recently submitted motions to dismiss five out of the six counts lodged in the lawsuits. In this article, we delve into OpenAI’s defense strategy and explore the broader implications of copyright law and technological advancement.
OpenAI argues that the transformative nature of its LLM technology should be taken into account when assessing the copyright claims. The company emphasizes the importance of striking a balance between copyright protection and technological progress. OpenAI contends that its LLMs, particularly ChatGPT, have immense value and potential for enhancing productivity, simplifying daily tasks, and aiding in coding. Drawing an analogy to the invention of the printing press, OpenAI suggests that ChatGPT represents a significant intellectual revolution.
Gregory Leighton, a privacy law specialist at law firm Polsinelli, notes that OpenAI’s defense highlights the limitations of copyright. He points out that processing copyrighted works in an LLM does not necessarily constitute a derivative work. Leighton suggests that copyright protection does not extend to facts and ideas, and OpenAI’s argument aligns with this perspective.
OpenAI defends its use of copyrighted materials by highlighting the fundamental characteristics of LLM technology. LLMs are neural networks trained on extensive text data to effectively comprehend human language, enabling users to input text prompts and receive generated content. According to OpenAI, its products merge LLMs with parameters that ensure the accuracy, relevance, safety, and utility of the generated outputs.
The plaintiffs argue that ChatGPT was trained using their copyrighted works without permission. OpenAI, however, asserts that fair use exceptions, an integral part of copyright law, can accommodate transformative innovations like LLMs. OpenAI argues that fair use promotes scientific and artistic progress, thereby aligning with the constitutional intent of copyright law.
Leighton points out an interesting aspect of OpenAI’s defense strategy. He argues that fair use is typically an affirmative defense and should not be raised in a motion to dismiss. The burden of affirmatively pleading and proving fair use lies with the defendant. OpenAI’s motion cites court cases where the fair use doctrine protected innovative uses of copyrighted materials, urging the dismissal of secondary claims by the plaintiffs.
Dissecting the Plaintiffs’ Claims
OpenAI challenges the legal validity of the plaintiffs’ secondary claims, such as vicarious copyright infringement, violations of the Digital Millennium Copyright Act (DMCA), violations of California’s Unfair Competition Law (UCL), negligence, and unjust enrichment. The company contends that these claims are based on flawed legal reasoning and should be removed from the lawsuit.
Vicarious copyright infringement is applicable when a party indirectly benefits from copyright infringement committed by another person. OpenAI argues that the allegations of direct infringement by the plaintiffs are not valid as a matter of law. The company refutes the notion that it had the “right and ability to supervise” or any direct financial interest in the alleged infringement.
OpenAI also disputes the plaintiffs’ claims that ChatGPT outputs are infringing derivative works of their copyrighted books. The company insists that the plaintiffs lack evidence to substantiate their claims and suggests that applying such standards on a wider scale could have unintended consequences for various industries. For instance, painters referencing photographers’ material could potentially face copyright infringement lawsuits.
Furthermore, OpenAI counters the plaintiffs’ arguments regarding the removal of “copyright management information” through LLM training. The company asserts that the plaintiffs’ evidence on this matter is contradictory and fails to demonstrate intentional removal.
OpenAI also challenges the negligence and unjust enrichment claims. The company claims that there are no grounds for negligence, as any use of copyrighted material by OpenAI or its users would be intentional. Additionally, OpenAI argues that it does not owe the plaintiffs a duty of care. The company rebuts the plaintiffs’ claim that OpenAI held onto profits or benefits derived from the infringed material, contending that no evidence supports such allegations.
Finally, OpenAI maintains that state law claims of negligence and unjust enrichment are preempted by federal copyright law, bolstering its argument for the dismissal of these claims.
Implications for the Future of AI and Copyright Law
While the immediate effects of OpenAI’s motion to dismiss may be limited, the outcome of these lawsuits holds significant implications for the intersection of AI and copyright law. If it is determined that training AI models on copyrighted works without permission always infringes copyright, it could impact various AI use cases.
Leighton highlights that these motions offer a glimpse into OpenAI’s overall defense strategy and provide insight into the complex legal landscape they must navigate. He notes that the response from the plaintiffs, which will outline their arguments for keeping the claims intact, will be an important development in the case.
As this legal conflict unfolds, it is likely to shape the future of copyright law and technological progress. Balancing the rights of content creators, the potential for transformative technologies, and the advancement of AI will require careful consideration and legal interpretation.
OpenAI’s motions to dismiss the copyright lawsuits shed light on the transformative nature of its LLM technology and its reliance on fair use principles. The company argues for the dismissal of secondary claims and challenges the legal validity of the plaintiffs’ allegations. The court’s decision on these matters will provide key insights into the future of copyright law and its implications for AI development and usage.