Patentability of Artificial Intelligence on the Precipice of Reform

• Jurisdiction: United States,Federal
• Author: Hannah Shows, Ph.D.
• Citation: Vol. 48 No. 3
• Publication year: 2023

PATENTABILITY OF ARTIFICIAL INTELLIGENCE ON THE PRECIPICE OF REFORM

Hannah Shows, Ph.D.

Haynes Beffel & Wolfeld LLP

HISTORICAL CONTEXT FOR PATENTING AI

Rapid expansion of artificial intelligence (AI) technology was a defining theme of 2022 and it continues to dominate market trends in 2023. Adoption of AI has more than doubled in the previous five years, as highlighted by extensive media coverage of generative AI tools like ChatGPT and DALL-E, drug discovery advancements thanks to protein AI AlphaFold, and controversy surrounding the increasing autonomy of self-driving cars. The current AI wave began approximately a decade before the recent boost in popularity—the number of academic papers relating to AI published annually increased six times from 1996 to 2010,¹ and the total number of AI papers published has since doubled.² Patent practitioners are undoubtedly aware of these trends, given the nearly 350,000 filed patent applications for AI-related inventions by 2019 (notably, over half of which were filed since 2013).³

The surge of patent filings related to AI has been accompanied by considerable challenges associated with subject matter eligibility, inventorship, and litigation. As supported by the 2019 Revised Patent Subject Matter Eligibility Guidance (2019 PEG) clarification of the concept of the abstract idea, AI-related inventions can be computer-implemented technology that is patent-eligible under 35 U.S.C. § 101, provided that the invention falls under one of the four statutory categories and is eligible under the Alice/Mayo⁴ test. The effects of the Alice decision⁵ undeniably linger in the form of uncertainty amongst inventors, practitioners, and examiners alike regarding the patentability of claims directed to an AI technology despite further USPTO guidance on the prosecution of AI-related patent applications and added clarification on software patent eligibility from the 2016 Federal Circuit Decision in Enfish v. Microsoft.⁶ Moreover, once an AI patent is granted, complex questions remain within AI-related patent litigation. If a first entity develops an AI and a second entity uses the AI to develop a product, and the product infringes on one or more claims of an issued patent, who is the infringer? How does the "black box" nature of many AI technologies (i.e., the process

[Page 68]

by which the AI generates an output is not readily interpretable) influence the identification of direct infringement? How feasible is it to detect and prove infringement in AI cases? These are only a small subset of the unique infringement challenges presented by AI that must be carefully considered when drafting and prosecuting AI-related patent applications to prioritize enforceability.

Ambiguity surrounding the prosecution and enforcement of AI-related patents is not unique to the US patent system. Given the widespread growth of the global AI market and competition from other countries' thriving tech sectors, it is unsurprising that AI developers frequently invest in foreign patent prosecution. In 2019, per the WIPO Technology Trends Report on Artificial Intelligence⁷ one out of every three AI patent applications are filed in additional jurisdictions after their first filing and eight percent are filed in five or more jurisdictions. Patent practitioners are well-equipped to be cautious of the potential effects foreign prosecution may have on the future litigation of a related application. Each respective patent jurisdiction is associated with its own unique guidelines, best practices, and challenges relating to the patentability of AI. Hence, patent practitioners and inventors are tasked not only with the navigation of patent laws in various states of flux within respective countries, but also the varying degrees of differences between them and the potential risks associated with differing prosecution strategies for different patent offices.

Given the rapidly evolving and varied nature of the AI patent landscape, it can be difficult for practitioners and inventors to access comprehensive information within patent law for AI inventions. The recognition and treatment of AI-generated inventions significantly varies across various jurisdictions, posing considerable challenges for AI developers in the modern globalized economy. Accordingly, an international consensus on future legal reform is highly advantageous.

DABUS AND AI-GENERATED INVENTIONS

Speculation upon the future of the AI patent landscape necessarily involves a summary of the DABUS saga currently prompting unique questions relating to AI inventorship.⁸ Decisions by the respective courts for the fate of DABUS will set crucial precedent for the patentability of AI-generated inventions. In recent years, a number of patent applications claiming AI-generated outputs have been filed by and assigned to Stephen L. Thaler, Ph.D., with support from a team of patent attorneys within a collaboration titled the Artificial Inventor Project.⁹ The applications filed by Thaler (collectively referred to by the name of the AI system, DABUS) name the AI system, rather than Thaler himself, as the inventor.

Thaler is a computer scientist and AI expert with over forty years of experience in neural network and deep learning-based research. Thaler is the creator of an AI paradigm titled Device for the Autonomous Bootstrapping of Unified Sentience (DABUS) configured to generate novel output patterns using a combination of neural network, reinforcement learning, and bootstrapping methodology. Thaler has applied for copyright and patent protection related to inventions generated by DABUS with a myriad of patent offices, including the USPTO. The listed inventor for the various DABUS patent applications is "DABUS (Invention generated by artificial intelligence)".¹⁰ The DABUS applications received nearly universal rejection internationally on the grounds that DABUS is not a natural person.¹¹Thus, the notoriety gained by DABUS is not in response to the patentability of AI-generated works, but to the question of whether an AI could be named as an inventor in a patent application. In fact, Thaler has already been granted almost two dozen patents relating to DABUS and similar technology.¹²

US Pat. App. No. 16524350 titled "Devices and Methods for attracting attention" was filed July 29, 2019, assigned to Thaler, and listed DABUS as the inventor. Thaler filed a substitute statement under 37 C.F.R. 1.64 in lieu of the declaration

[Page 69]

and a statement under 37 C.F.R. 3.73(c) for the assignment. Thaler executed the documents on behalf of DABUS, as legal representative of the assignor and on behalf of himself as the assignee. A Notice to File Missing Parts was issued by the USPTO on August 8, 2019, indicating that the ADS did not identify the inventor by their legal name. Applicant petitioned to request vacation of the Notice and the request was ultimately dismissed. Applicant took the case to the District Court and to the Court of Appeals of the Federal Circuit. The Court of Appeals affirmed the USPTO decision by confirming that an inventor must be a human, stating in its decision that "Congress has determined that only a natural person can be an inventor, so AI cannot be."¹³ Applicant appealed to the Supreme Court on March 17, 2023 to answer the question of whether the Patent Act categorically restricts the statutory term "inventor" to human beings alone. Most recently, the Supreme Court of the United States declined to hear the appeal on April 24, 2023.

While the DABUS cases serve to specifically address whether patent law categorically restricts the statutory term "inventor" to human beings alone, the impact of the decisions reaches much further. All decisions relating to AI inventorship carry immense consequences for AI patentability on a broad scale.

If the decision that AI cannot be an inventor for the purpose of obtaining a patent stands, but the owner of the AI also does not qualify as the inventor if they did not contribute to the conception of the...