First published by ALM / Law.com in The Intellectual Property Strategist

All of us have been exposed to and perhaps even overwhelmed by news about generative artificial intelligence (AI). Unlike machine learning technology that merely classifies or predicts, generative AI creates. Industry stalwarts and startups alike have launched generative models that can create new text, images, video, 3D models, and even software code — with the promise of more powerful and disruptive innovations to soon follow. A patent strategy informed by the unique considerations raised by generative AI will optimize protections for innovations in the field. Patent strategies should reflect the current legal landscape as well as anticipate potential future legal developments.

Part One of this article series covers claim scope and inventorship. Part Two of this article series covers subject matter eligibility, prior art, and future opportunities.

PART ONE

Patent Claims and Inventorship 

The Federal Circuit in Thaler v. Vidal, No. 2021-2347 (Fed. Cir. 2022) recently confirmed that an inventor under the patent statute must be a natural person. In this case, Thaler filed with the U.S. Patent Office two patent applications in which artificial intelligence was identified as the inventor. Without regard to the nature of the invention, the Federal Circuit categorically rejected this characterization of inventorship. The decision underscores human innovation as requisite for patent protection in the U.S. The Supreme Court has just refused to hear the case.

Subject to legal and regulatory developments under active consideration by policymakers, patent strategies relating to generative AI should heed Thaler. Current attempts to claim only the output of a generative model could implicate non-human “inventorship” (and of course other issues such as printed matter) and thus lead to patent futility. A more comprehensive claim strategy will likely yield better results. In this regard, wider focus on training, design, tuning, or application of generative models will better satisfy inventorship requirements to preserve the possibility of patent rights. For example, conduct the disclosure meeting to explore design choices and configurations for a generative model that were selected through human effort. As another example, proceed beyond the generative model output to cover applications or adaptations for which the generative model will be most likely or valuably utilized. Claims scope that extends to cover components of a system incorporating a generative model may be especially relevant when, as current trends indicate, the generative model, whether closed or open sourced, is not homegrown but rather leveraged through a third-party API. Patent claims directed to these considerations, and not to a generative model output alone, cover human innovation, implicate inventorship by a natural person, and likely optimize a path to patent protection.

The requirement of human inventorship is not universal. The Federal Circuit in Thaler expressly acknowledged that South Africa, which has a patent system with legal requirements and standards different from the US patent system, granted patents for the same patent applications. As generative AI increases in importance around the world, patent systems in other jurisdictions will likely need to also confront and decide the boundaries of inventorship required for patent grant. For example, so far the European Patent Office has required a natural person and the UK Supreme Court is expected to issue a ruling on the issue. As the legal landscape develops, legal requirements in regions of interest relating to generative AI inventions should be periodically reassessed to inform the geographic scope of patent efforts. Companies should know when a patent grant is technically impossible in a particular region due to a failure to meet a local inventorship definition. Otherwise, an uninformed patent application filing in the region and subsequent publication would not only fail from a patent rights perspective but also effect a needless forfeiture of other possible intellectual property rights (e.g., trade secret rights) in the invention.