Safeguarding Human Invention in the Age of Artificial Intelligence
![](https://ipleadership.org/app/uploads/2023/11/GettyImages-1366475366_Crop1.jpg)
By Alexander Kersten
Introduction
On October 30, President Joe Biden issued the Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence. This comes at a time when artificial intelligence (AI)’s anticipated capabilities and applications are expected to have a tremendous impact on the future of U.S. competitiveness and national security. One area of concern relates to China’s ability to produce leading-edge chips that can enhance its AI capabilities. Recognizing this vulnerability, the United States in October 2022 announced unprecedented export controls targeting China’s military AI development by limiting its access to advanced semiconductors and the tools used to manufacture them. These controls were further refined on October 17, 2023.
Amid a contentious debate further muddied by misunderstandings about the technology’s current capabilities and fears that bad actors would weaponize it, this latest executive order has been noted most for its recommendations that AI firms conduct safety tests and share those results with the government, that standards and industry norms be established for this emerging area of critical technology, and that government agencies implement stricter changes in how these use AI.
Less discussed, however, has been the executive order (EO)’s introduction of a policy framework to assess the question of AI and inventorship, as well as patentable subject matter. The White House’s own pre-released fact sheet did not mention these aspects of the order, meaning that the many articles and commentaries written since also fail to discuss this important development in how the United States treats the question of AI and inventorship. Recognizing the complex realities already facing patent examiners at the United States Patent and Trademark Office (USPTO), the order’s text states that the Under Secretary of Commerce for Intellectual Property and USPTO director must “within 120 days of the date of this order, publish guidance to USPTO patent examiners and applicants addressing inventorship and the use of AI, including generative AI, in the inventive process” following with further timelines and guidelines for the USPTO and Department of Commerce more generally.
Intellectual Property Law and the AI Revolution
This EO comes at a key time for the recognition of intellectual property (IP) rights associated with AI-generated creations and inventions. A 2020 USPTO report cites the agency’s 2019 request for comment (RFC) found that AI “has no universally recognized definition” with a wide range of public commenters, including individual stakeholders, associations, corporations, and foreign IP offices agreeing that “the current state of the art is limited to ‘narrow’ AI,” referring to AI programmed to perform single and often repetitive tasks, rather than capable of tackling issues related to the broader gambit of “artificial general intelligence” that would by definition rival human intelligence, perhaps with unsettling consequences. While popular predictions and media depictions polarize opinions on the impact of AI, the truth about how it is used to bolster human creativity is more nuanced. Sensitive policymaking requires looking beyond the hype on one hand and the doom and gloom on the other.
AI technology has piqued the interest of IP professionals, not only because of how it impacts the pace of innovation, but also because of challenges that the rapid evolution of AI platforms it presents to the IP regime—particularly in terms of ensuring security, confidentiality, and trust in the system. AI also raises key questions related to independent creativity and originality of ideas and their ownership. While the world has indeed entered a profound new era where a machine can apparently create something on its own, how autonomously can it do so? Does it recognize the utility of its creation, and how much human intervention is still needed to ensure that process? Can AI systems actually “invent,” how would humans ever compete with them, and should they even be entitled to rights over their inventions?
These puzzles in turn prompt the need to see if the current system of IP rights can handle new capabilities posed by AI-aided invention. In looking at instances of AI use in the invention creation process, especially in the biopharmaceutical sector, a joint CSIS project with the Special Competitive Studies Project in consultation with leading legal and IP experts reveals that current patent law requiring a patent application to recognize a human inventor should indeed be sufficient to address the role of AI with its current capabilities in the invention creation process.
How AI-Aided Inventions Are Challenging the Patent System
The U.S. patent system is widely regarded as an important economic tool for incentivizing innovation. As such, it is a key foundation of the modern economy. By codifying an inventor’s ownership of their inventions for a time, inventors are encouraged to divulge their discoveries to the public domain, encouraging cooperation in research, development, and commercialization and driving further innovation.
The U.S. patent system is rooted in the U.S. Constitution, as outlined in Article 1, Section 8, Clause 8. By patenting a new product, for example, the inventor is granted rights over their IP for a specified time during which the inventor can profit from the invention exclusively. To receive a patent, a would-be inventor must sufficiently prove that their invention is something legally patentable by demonstrating to patent examiners that the product has utility and novelty, is “nonobvious” (something that is not apparent to someone with a background in that field or “prior skill in the art”), and has “enablement”—meaning a description of the invention and how it can be made or replicated and with what processes or materials, given that the replicator is similarly “skilled in the art.” Throughout U.S. history, technological “revolutions” and waves of new inventions have tested this framework, and the rise of AI represents the latest challenge.
Current U.S. law is quite clear that an inventor must be a human. Yet the USPTO has been challenged in recent years by patent applications from an individual named Stephen Thaler, filing in 2021 on behalf of an AI software named Device for the Autonomous Bootstrapping of Unified Sentience (DABUS). Having filed applications in the United States, European Union, United Kingdom, Japan, New Zealand, Australia, and South Africa, Thaler faced IP regimes of varying conceptions across these countries regarding inventorship by a nonhuman entity.
In Thaler v. Vidal the USPTO concluded that DABUS did not qualify as an inventor because it does not meet the definition of an individual (natural person or human) under the current law 35 U.S. Code § 115 – Inventor’s oath or declaration. Further, the decision pointed out that to be an inventor under current law requires having “intentionality and culpability,” which AI does not have because it cannot legally enter into contracts. However, the Thaler case did not specifically examine the specific contexts as to where a human uses AI as a tool, where AI contributes to invention-creation, and if those inventions are patentable.
While the United States, European Union, United Kingdom, New Zealand, and Japan did not recognize a nonhuman as an inventor, both Australia and South Africa awarded patents to DABUS for the inventions of a novel food container and a light-up stick. In this same vein, questions over AI’s potential to be listed as a co-inventor in the United States with a human are settled by AI’s inability (so far) to recognize the utility of an invention, failing to fulfil a co-inventor’s requirement that “when an invention is made jointly, the joint inventors need not contribute equally to its conception” as laid out in the 2018 opinion In re Verhoef, citing 35 U.S. Code § 116.
AI Is a Sophisticated Tool for Human Inventors
With the recent launch of generative AI models, many inventors, investors, and others are exploring and testing AI’s capabilities. AI remains a sophisticated tool to expand human creativity, but is not yet capable of independent inventorship or conception. Accordingly, the patent system should continue to support human innovators that exploit AI capabilities.
In the process of drug discovery in the biopharmaceutical sector, for example, AI may be supplied parameters, but it is ultimately highly trained human scientists who must set those parameters and go through the technical and laborious process of testing the AI’s results in a laboratory before being able to assess the safety, utility, or value of anything that the AI has generated.
In short, the current use of AI in highly specialized research settings can drive gains in efficiency and churn out novel thinking, but analyzing the practicality of these results still calls for skilled human interpretation—where the human in this example would be clearly eligible as an inventor under U.S. patent law. Specifically, as AI’s use in similar processes grows more sophisticated, those employing the use of AI will require guidance on which “individual” working with the AI qualifies as the inventor.
Protecting the Role of the Human Inventor in the AI Era
As AI becomes more ubiquitous and makes greater contributions to the invention process, the role of protecting and promoting human inventorship becomes critical. Humans should continue to be recognized as the actual inventors, even as these inventors use AI as a “sophisticated tool” of experimentation and discovery.
President Biden’s executive order is right to note that the Department of Commerce and USPTO have a responsibility to make this distinction explicitly clear to stakeholders. By “clearing the air,” this distinction will help ensure the widespread use of increasingly advanced AI tools in the invention process. Investors and innovators need clarity that use of AI in the invention creation process will not lead to the invalidation of patents down the line. Guidance will be needed on edge cases regarding what does and does not constitute contribution to an invention.
The accelerating progress of AI stands to make ever more significant contributions to the inventive process, blurring the demarcation of inventive contribution. With progress will come new questions: Should AI ever be regarded as a joint or even sole inventor? Should there be IP protection for the datasets used to train artificial intelligence machines? And what are the patentability standards? Under current law, several standards revolve around knowledge of those of “ordinary skill in the art”—but who are those of ordinary skill when machine minds are also considered?
Through this turbulence, the United States will need an intellectual property regime that provides certainty and concrete guidance. The Department of Commerce and the USPTO should establish procedures that enable the study of patent regime revisions as and when the need appears. Although the specifics of these updates will only become clear as those capabilities and applications emerge, the United States should be ready to swiftly adapt to an abrupt paradigm shift in AI inventorship and other pressing issues.
Alexander Kersten is the deputy director and fellow with the Renewing American Innovation Project at the Center for Strategic and International Studies in Washington, D.C.
This piece was originally published on August 28th, 2023 with the Renewing American Innovation (RAI) Project at the Center for Strategic and International Studies (CSIS)
![](https://ipleadership.org/app/uploads/2024/08/BiopharmaStockPhoto-800x450.jpg)
![](https://ipleadership.org/app/uploads/2024/07/AdobeStock_210923686_cropped-800x450.jpeg)
Explainer: Standard Essential Patents and National Security
![](https://ipleadership.org/app/uploads/2024/06/GettyImages-483223361_cropped-800x450.jpeg)