When Innovation Creates: Additional Developments in Artificial Intelligence at the U.S. Patent and Trademark Office

AI systems are already creating works that, if created by a human, would warrant IP protection. Reports of AI-created art, music, literary works, data, technology, formulas, flavors, and other innovations are becoming increasingly common. In 2016, for example, a new "Rembrandt" portrait was unveiled in the Netherlands, generated by an AI system that analyzed more than 300 real paintings of the Dutch master and then used AI, facial recognition, and 3D printing technologies to create an entirely new work in the same style. In 2018, a portrait created using AI sold for $432,500 at auction. A novel publicized as being the first book written by AI was published last year.

At the same time, interesting questions arise as to whether and to what extent AI's disruption of the retail experience will impact traditional concepts of trademark law. Consumers' purchasing decisions are increasingly influenced by AI, as hyper-personalized advertising and predictive retailing have become commonplace. If a product or service is suggested to us by an AI system based on our demographic profile and previous shopping habits, then can it be liable if it suggests infringing or counterfeit goods? Novel questions also arise in the trade secret context. Can a company deploy an AI system to develop trade secrets and, if so, can the AI adequately explain what that secret is and maintain its secrecy? When is it acceptable for an AI system to try to discover another party's secrets?

There can be no doubt that AI is disrupting fundamental notions of authorship and inventorship of IP and raising serious questions about whether traditional forms of IP protection are sufficient to protect AI-created innovation. Indeed, the Copyright and Patent Clause of the U.S. Constitution recognizes the power of Congress to give "authors" and "inventors" exclusive rights for a limited time period—and, until recently, it was generally understood that authors and inventors were human beings. In 2014, in response to the infamous "monkey selfie" case, the Copyright Office clarified that human authorship is indeed a requirement for copyright protection, noting that a "photograph taken by a monkey" is not copyrightable. However, in language that has implications for AI-created works, the Copyright Office also noted that "works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author" are also not copyrightable. However, what remains unclear—and likely the source of future debate—is what level of human input or intervention is necessary to be eligible for copyright protection.

As we previously reported, recent patent applications naming AI as the inventor have accelerated discussions about whether and to what extent (and under what circumstances) AI can "invent" a patented technology, as well as a variety of other questions related to the impact of AI on patent law and policy. Shortly after these applications were filed in August, the USPTO issued a notice seeking public comment on a variety of questions related to AI in the patent context, with the deadline for such comments extended to November 8, 2019.

Now, the USPTO has extended its inquiry to the impact of AI on nonpatent IP rights, including copyrights, trademarks, and trade secrets. Issued on October 30, the USPTO's notice broadly requests input on any issue related to AI and these various IP rights, but also enumerates a number of specific questions. With respect to copyright, the USPTO asks whether a work produced solely by an AI algorithm or process should qualify as a work of authorship protectable under U.S. copyright law (and, if human involvement is necessary, what level of involvement is sufficient), whether existing law (e.g., the fair use doctrine) adequately addresses the legality of an AI system's use of copyrighted material to learn (i.e., train a machine learning model, etc.), whether current law adequately addresses the scenario in which an AI system creates a work that infringes a copyrighted work, and whether an entity can claim ownership over a work created by AI when that entity trained the AI process. The USPTO also asks if using AI in trademark searches would impact the registrability of trademarks, how AI affects both trademark and trade secret law, whether existing laws are adequate to protect databases and data sets, and whether specific IP-related practices from other countries might be helpful in informing the USPTO's policies moving forward. The deadline for submissions is December 16, 2019.

What effect the USPTO's solicitation of comments will have on copyright and other IP protections for works produced by AI remains to be seen. However, given the USPTO's ongoing interest in the impact of AI on IP rights, it seems likely that the USPTO will offer additional guidance on the topic. Stakeholders should watch these developments closely, particularly as AI-related IP issues are debated among lawmakers and tested in the courts.