Last month, the Supreme Court handed down a ruling that had nothing — and everything — to do with AI: the Cox Communications v. Sony Music Entertainment decision. By siding with an internet provider in a billion-dollar music piracy case, the Court quietly built a fortress around the future of AI innovation.
The Court’s decision reduces copyright liability exposure for AI companies based on infringement by their users. That outcome has enraged entities that create copyrighted material, such as movie studios and record labels, but it keeps in place fair use protections for dual-use technologies traditionally recognized by copyright law. Thus, the Supreme Court has extended a significant zone of protection around AI innovation.
Cox is a broadband internet service provider, and like all ISPs, some of its customers violate copyright law by sharing sound recordings. Sony sued Cox, rather than its customers, on the theory that Cox was indirectly responsible for the infringing activities of its users — a doctrine known as contributory infringement in copyright law. Sony Music won at trial and was awarded more than $1 billion in damages.
The Supreme Court reversed the verdict in Sony’s favor, clarifying that service providers such as Cox are contributorily liable for their users’ infringement only if the provider intends that its service be used for infringement. Intent, the Court held, can be shown in two ways: either that the provider’s service is tailored to infringement — it has no commercially significant or substantial non-infringing uses — or that the provider actively induces its customers to infringe, such as by advertising infringing uses or building a business model dependent upon unauthorized file sharing.
Cox did not fit either variant: Internet access can be employed for a wide range of activities, lawful and unlawful, and the ISP had better sense than to suggest infringement to subscribers. Accordingly, Cox is not contributorily liable for infringement by its users, and online service providers breathed a sigh of relief.
From a copyright law perspective, AI foundation model providers bear a striking resemblance to Cox — a resemblance that may well have been on the justices’ minds as they crafted the decision. Like internet access services, AI foundation models are classic dual-use tools: they can be employed for conduct that respects copyright protection or violates it, depending on the behavior of a particular user. For example, generative AI systems now perform many functions that are recognized as fair use, such as enabling search functionality and research.
Applications of AI enhance the socially beneficial applications of fair use. Because of these capabilities, most AI systems are not tailored to infringement, protecting them from liability. Historically, when regulating dual-use technologies, copyright law has concentrated on the conduct of consumers, not producers, and that focus sensibly balances protection for both creative expression and innovation.
Also, generative AI app developers have learned caution in how they market such tools after the controversy generated by the debut of OpenAI’s Sora 2 video generation system, which initially enabled users to generate infringing content unless copyright owners specifically opted out. OpenAI pivoted quickly to a more restrictive policy, and many developers have implemented filters to block infringement by users. Anthropic’s Claude will refuse to create a Donald Duck image, for example. These safeguards demonstrate that AI developers are not tailoring their systems for infringement, nor are they encouraging users to violate copyright law.
After the Cox decision, it will be quite difficult for copyright owners to show that providers of online AI services intend for their users to infringe. But this is beneficial because dual-use technologies create immense social value, facilitate socially beneficial fair use, and promote a thriving commons that benefits consumers and creators. Technologies such as the photocopier, VCR and even the internet faced legal challenges based upon copyright, but the social value of these advances is enormous.
Similarly, copyright infringement from the use of AI systems does not yet pose any real financial risk to content owners. Even if Sora 2 enables users to create infringing clips that copy from SpongeBob Squarepants or Pokemon, those clips do not displace sales by Nickelodeon or Nintendo.
If someone does start to earn revenue from AI-generated SpongeBob episodes, Nickelodeon can simply sue the creator to recover damages or block distribution via an injunction. The benefits from AI systems come with little to no financial cost to content owners; where there is no harm, there should be no liability.
In effect, the Supreme Court has issued its first AI opinion with the Cox decision. By protecting dual-use technologies while allowing copyright owners to sue firms that are designed for or marketed to infringement, the Court has maintained a careful balance that simultaneously promotes technological innovation and creative expression.
About the authors:
Derek E. Bambauer is the Irving Cypen Professor of Law at the University of Florida Levin College of Law. A former principal systems engineer at IBM, his research concentrates on AI, cybersecurity and intellectual property.
Thinh N. Nguyen is director of the Innovation and Entrepreneurship Clinic at the University of Florida Levin College of Law. He is a nationally recognized expert on law and AI with more than two decades of experience representing technology firms.
