
Blurring the Line Between the Dry and Wet Lab: Joint Inventorship in AI-Assisted Life Science Inventions
In 2024, not one but two Nobel Prizes (in Chemistry and Physics) were awarded to researchers for their work in artificial intelligence ("AI"). Particularly noteworthy for the life science community is the Nobel Prize in Chemistry, awarded to David Baker for "the almost impossible feat of building entirely new kinds of proteins" and to Demis Hassabis and John M. Jumper for "develop[ing] an AI model to solve a 50-year-old problem: predicting proteins' complex structures." These feats cement AI's place in the pantheon of life science breakthroughs and highlight AI's increasing role in biopharmaceutical research and development ("R&D"). In recent years, scientists have demonstrated the viability of AI in nearly every stage of the drug discovery pipeline, from druggable target identification to lead molecule screening and optimization to clinical trial design. However, despite promises to increase the speed and success rate of drug development, the use of AI challenges traditional views on a fundamental tenet of patent law: inventorship.
Under U.S. law, an AI cannot be an inventor on a patent. The Federal Circuit's 2022 Thaler[i] decision expressly considered this issue and held that the U.S. Patent Act defines "inventor" as limited to human beings, thus precluding non-natural persons (including AI) from being listed as inventors on U.S. patents and patent applications. The "human inventorship" requirement for patents is shared by most countries that have considered the issue (with the exception of South Africa). Notably, however, Thaler did not consider whether inventions made by humans with the assistance of AI are eligible for patent protection. Instead, this issue became the focal point of the Inventorship Guidance for AI-Assisted Inventions ("inventorship guidance" or "guidance"), 89 Fed. Reg. 10043, released by the United States Patent and Trademark Office's ("USPTO" or "the Office") on February 13, 2024.
According to the USPTO's inventorship guidance, while AI-assisted inventions are not per se unpatentable, at least one natural person must make a significant contribution to that invention in order for there to be proper inventorship. Other jurisdictions, such as by the European Patent Office and China National Intellectual Property Administration, also impose similar requirements for human contribution. Nevertheless, no bright-line test exists to gauge what constitutes a "significant contribution." Even in the absence of AI, the statutory framework and case law for inventorship are murky—the proliferation of AI in the R&D process adds further uncertainty and complexity. For example, under the guidance, the scope of "significant contribution" is expanded to include contributions made by humans through interactions with an AI system that is an "essential building" block of an invention. With several notable new grounds to support an inventorship claim, heightened inventorship scrutiny may be especially critical in biopharma and other open-innovation industries where inventions are often cumulative and include contributions from multiple entities. To better understand the implications of the USPTO's inventorship guidance for the life science industry, we examine the inventorship of AI-assisted inventions in the context of the inclusive inventorship standard that evolved to promote an open-innovation paradigm.
Proper Inventorship Is Fundamental to the Validity and Enforceability of a Patent
Patents are a negative right, meaning that the owners of a patent are afforded the right to exclude others from practicing the claimed subject matter of the patent. This negative right gives rise to a limited monopoly, an often lucrative reward intended to incentivize the sharing of ideas and encourage investments in R&D. Patents are especially valuable assets in the life science industry, which relies on the exclusivity afforded by patents to recoup the many billions of dollars poured into the R&D of pharmaceuticals. Inventorship is a fundamental tenet of patent law at least because under U.S. law, inventors are the presumed owners of a patent and are thus the beneficiaries of the concomitant patent monopoly until a transfer of some or all of those rights occurs. A patent with improper inventorship can be an expensive mistake. For example, omitting a person who should be named as an inventor or naming someone who does not qualify as an inventor can result in the patent being declared invalid and unenforceable. Failure to properly account for every possible inventor can also complicate IP ownership when the inventors are affiliated with different entities. Especially pernicious are situations in which an unwitting collaborative relationship gives rise to joint ownership, thus undermining the exclusivity of the consequent patent assets.
Joint Inventorship and the Open-Innovation Industry
Inventorship is a legal concept that can at times defy the norms of scholarly collaboration and academic authorship observed by the scientific community at large. The USPTO's inventorship guidance for AI-assisted inventions leans heavily on the existing legal framework for human inventorship. First, the guidance applies the Pannu factors, which define an inventor or a joint inventor as a person who makes a contribution to the conception of an invention that is not insignificant in the context of the full invention and transcends mere explanation of the current state-of-the-art.[ii] Notably, conception is defined as the mental part of the inventive act. It is not the mere recognition of a problem or having a general goal. Instead, conception requires formulating the means of achieving a particular end result or goal. Conception is therefore distinct from the concept of "reduction to practice" or the physical act of completing an invention for its intended purpose. However, in limited circumstances, it is possible for conception to occur simultaneously with reduction to practice, meaning that the individual(s) who reduced an invention to practice can also qualify as inventor(s) and not just those who conceived of the idea. The doctrine of simultaneous conception and reduction to practice is especially germane in the context of life science R&D where the utility of an invention, such as a drug molecule, is often uncertain until it undergoes synthesis and testing
Second, the guidance also relies on the standard for joint inventorship codified under 35 U.S.C. §116(a), which was amended in 1984 to relax the collaboration requirement for joint inventorship.[iii] In particular, this requires only a modicum of collaboration between joint inventors and, importantly, does not require joint inventors to work together physically or at the same time and does not require each joint inventor to have contributed to the subject matter covered by all of the claims in a patent application.[iv] Moreover, to the extent each joint inventor is required to make a significant contribution, there is in fact no minimum threshold for the quantity or quality of contribution.[v] What constitutes "significant contribution" remains a fact-specific inquiry, as the courts have yet to prescribe any bright-line rule.
The result is an inclusive inventorship standard that reflects a concerted effort by Congress and the Federal Circuit to develop an open-innovation-friendly inventorship paradigm.[vi] The cumulative nature of inventions in open-innovation industries means that those industries are especially susceptible to so-called "secret" prior art.[vii] Despite not being publicly accessible, secret prior art, such as unpublished patent applications and surreptitious offers for sale, can still be used to challenge the patentability of subsequent improvements. This is true both under the first-to-invent regime that was in place before the America Invents Act ("AIA") and the current post-AIA first-inventor-to-file regime. The common ownership exception introduced by the same 1984 amendments was not an adequate solution because R&D in open-innovation industries often span multiple entities. Thus, lowering the bar for joint inventorship is thought to be more open-innovation friendly because more prior art can be removed from consideration under the common inventorship exception. In other words, an inclusive joint inventorship standard increases certainty in the patentability of inventions that are products of open innovation. An inclusive joint inventorship standard is also a more flexible remedy than the joint research agreement exception established under the Cooperative Research and Technology Enhancement (CREATE) Act because no formal agreement is necessary to invoke the joint inventorship exception.
The AIA, in attempting to balance the interests of independent inventors who lack the resources to file patent applications early and often against those of large corporations who prefer greater certainty in patentability, retained at least some forms of secret prior art. For example, in Helsinn, the Supreme Court unanimously held that the on-sale bar against patentability under 35 U.S.C. §102 contemplates offers made in secret. As such, one could argue that there are still sufficient incentives to maintain an inclusive joint inventorship standard. However, the price for greater certainty in the patentability of an invention is more uncertainty in its inventorship. For instance, in the life science industry, innovative activities can range from pre-invention contributions, such as discoveries of laws of nature, to post-invention contributions, including experimentation to confirm drug efficacy.[viii] Any researcher participating in an open exchange of ideas and data can claim joint inventorship if that person's ideas are used to complete an invention.[ix] More recent cases have continued to embrace an inclusive joint inventorship standard, with the threshold for collaboration between joint inventors being satisfied even when a joint inventor is unaware of the invention being conceived.[x] The AI-assisted inventorship question is likely to add further uncertainty and complexity. While it is clear that AI itself cannot be a joint inventor, the individuals who can qualify may extend well beyond what a typical patent applicant in the life science field has been conditioned to expect.
USPTO's Guiding Principles Help to Determine Inventorship of AI-Assisted Inventions
The USPTO's inventorship guidance sets forth five guiding principles to help determine the inventorship of AI-assisted inventions. In particular, these guiding principles acknowledge that a natural person can make a sufficiently significant contribution to an AI-assisted invention by interacting (e.g., designing, building, and training) with an AI system. Below, we examine each principle in detail.
Guiding Principle 1. A natural person's use of an AI system in creating an AI-assisted invention does not negate the person's contributions as an inventor. In other words, a person using an AI system to create an initial design can qualify as an inventor if s/he makes a sufficiently significant contribution to the invention, for example, by modifying the initial design created by the AI system. Guiding principle 1 establishes that AI-assisted inventions are not categorically unpatentable for lack of proper inventorship. Instead, an AI-assisted invention can have proper inventorship if at least one natural person made a significant contribution to the subject matter of the invention.
Guiding Principle 2. Merely recognizing a problem or having a general goal or research plan to pursue does not rise to the level of conception. This aspect of principle 2 is consistent with the classic criteria for inventorship. More interestingly, principle 2 recognizes the possibility that a significant contribution can be made in the way humans interact with AI systems. Specifically, it is possible for an inventor to make a significant contribution by constructing a prompt for a specific problem in order to elicit a particular solution from the AI system. In the context of life science R&D, principle 2 introduces several possible joint inventors whose contributions arise from interactions with an AI system. Consider a fact pattern in which an AI system is used to generate new drug molecules by modifying an existing molecule provided as a prompt to the AI system. Under principle 2, the individuals who identified the existing molecule would qualify as inventors of the new drug molecules because by identifying the existing molecule, they effectively constructed a prompt for a specific problem and to elicit a particular solution from the AI system (e.g., generate new drug molecules with better properties than the existing molecule).
Guiding Principle 3. Reducing the output of an AI system to practice typically will not give rise to inventorship because conception is still the touchstone of inventorship. But where the utility of the output of the AI system is uncertain, a person who reduces the output to practice—for example, by conducting a successful experiment using an AI system's output —could demonstrate a significant contribution to the invention. Nevertheless, while the doctrine of simultaneous conception and reduction to practice is often associated with the so-called unpredictable arts, namely biology and chemistry, case law such as Burrough and Pannu indicate ample inconsistencies in when the doctrine is invoked. In practice, not every invention in the unpredictable arts is sufficiently unpredictable to invoke the doctrine of simultaneous conception and reduction to practice. Similarly, some inventions in the traditionally predictable arts, such as electrical, mechanical, and software engineering, may be sufficiently unpredictable to warrant the doctrine. Guiding principle 3 also raises the more fundamental question of just how predictable AI is itself. On the one hand, AI is often described as black box algorithms whose complex nonlinearities in extremely high dimensional spaces make it impossible to predict and explain its behavior. Accordingly, could the conception of at least some AI systems, and by extension its output, not occur until it is reduced to practice? On the other hand, AI systems are often designed and trained to produce certain outputs, such as drug molecules with specific properties like high binding affinity. Does this mean that the output of AI systems are reasonably predictable, such that the doctrine is less relevant in the context of AI-assisted inventions?
Guiding Principle 4. A natural person who develops an essential building block from which the claimed invention is derived may be considered to have provided a significant contribution to the conception of the claimed invention. In the context of AI-assisted inventions, this could be one or more individuals who designs, builds, or trains an AI system in view of a specific problem to elicit a particular solution in instances where the designing, building, and training of the AI system is a significant contribution to the invention created with the AI system. Guiding principle 4 is especially notable for its nod to a relaxed collaboration standard in which the individual(s) who designed, built, or trained the AI system can qualify as inventors without being present for, or act as a participant in, each activity that led to the conception of the claimed invention. It is also important to recognize that in real-life situations, multiple individuals, and by extension multiple entities, could participate in the development, building, and training of an AI system. For example, a first person could develop the architecture of a foundational model that is implemented by a second person and then trained using data curated by a third person to generate new drug molecules for synthesis and testing by an entirely different person altogether. Guiding principle 4 makes it possible for every one of these individuals to assert a claim of joint inventorship, even if very little actual collaboration transpired between them. Under principle 4, even individuals whose activities may seem far removed from what is considered core life science R&D can still qualify for joint inventorship through their interactions with an AI system.
Guiding Principle 5. Maintaining "intellectual domination" over an AI system does not, on its own, qualify a person as an inventor of any inventions created through the use of the AI system. In this context, intellectual domination refers to control over the AI system. Under principle 5, merely owning or overseeing an AI system used to create an invention without providing a significant contribution to the conception of the invention does not confer inventorship status. To the extent guiding principle 4 were to be construed to suggest anyone with control over an AI system is an inventor of inventions derived from the AI system's outputs, principle 5 undercuts that construction. However, principle 5 does not go as far as imposing a higher collaboration standard between the owners of an AI model and the individuals who use the AI model to create AI-assisted inventions. Instead, principle 5 applies the same criteria for assessing whether a "significant contribution" has been made by an owner of an AI system. It remains to be seen what acts beyond owning and overseeing an AI system constitute "significant contribution."
Best Practices
AI is causing a paradigm shift in life science R&D. One major ramification of this shift is the comingling of experts from traditionally disparate disciplines like software and hardware engineering and the life sciences. A growing number of people could be collaborating without realizing that the relationship rises to the level of joint inventorship under the prevailing inclusive joint inventorship standard. Consequently, inventorship should be a proactive inquiry that thoroughly examines every facet of the inventive process, from conception to reduction to practice. With AI opening new avenues for claiming inventorship status, it has become more imperative than ever before to understand and document the role of every contributor.
The inventorship guidance also affords USPTO personnel the ability to "request information … regarding inventorship even if the information is not material to patentability" in instances where "an examiner or other USPTO employee has a reasonable basis to conclude that one or more named inventors may not have contributed significantly to the claimed subject matter."[xi] In anticipation of such information requirements, patent applicants should be prepared to make the necessary disclosures without inadvertently revealing any trade secrets and other confidential information.
In sum, the USPTO's inventorship guidance contemplates a number of new bases for inventorship that should be considered by patent applicants whose inventions were assisted by AI. For example, although an AI system cannot be an inventor itself, it is conceivable that interaction with the AI system could be sufficient nexus for researchers and developers from different entities to qualify as joint inventors. Thus, to maintain clarity in IP ownership and avoid increasingly complex and costly disputes, inventorship should be monitored throughout the R&D process rather than relegated as a mere formality during patent procurement. Furthermore, the inventorship inquiry should consider not only human collaborators but also computational tools, and not necessarily only so-called AI tools, because what constitutes AI has and will continue to evolve. With the informal nature of some collaborative relationships, a thorough inventorship inquiry could be compromised by the passage of time. With significant R&D investments at stake, a "surprise" inventor from a different entity is a costly mistake that ultimately thwarts the ability to monetize patent assets. In another words, time is of the essence—it is a mistake to delay until the filing of a patent application to attempt to identify inventors.
Disclaimer
The USPTO published Inventorship Guidance for AI-Assisted Inventions, 89 Fed. Reg. 10043 (Feb. 13, 2024), under President Biden's order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, Exec. Order No. 14110, 88 Fed. Reg. 75191 (Oct. 30, 2023). President Trump revoked Executive Order 14110 and directed agencies to review all actions taken under Executive Order 14110 for compliance with a new "global AI dominance" policy. See Exec. Order No. 14148, 90 Fed. Reg. 8237, 8240 (Jan. 20, 2025) (rescinding Executive Order 14110 among others); Exec. Order No. 14179, 90 Fed. Reg. 8741, 8741 (Jan. 23, 2025) (announcing the new policy). But although President Trump directed agencies to "suspend, revise, or rescind … , or propose suspending, revising, or rescinding" actions that do not comply with the policy, he did not automatically revoke any such actions. 90 Fed. Reg. at 8741. To date, the USPTO's Inventorship Guidance for AI-Assisted Inventions remains in full effect and has not come under public scrutiny.
[i] Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022).
[ii] 89 Fed. Reg. 10046, 10048.
[iii] Id.
[iv] Kimberly-Clark Corp. v. Procter & Gamble Distrib. Co., 973 F.2d 911, 916-17, 23 USPQ2d 1921, 1925-26 (Fed. Cir. 1992) as cited in MPEP §2109.01.
[v] Toshiko Takenaka, Unravelling Inventorship, 21 Chi.-Kent J. Intell. Prop. 115 (2022).
[vi] Takenaka, pg. 71.
[vii] Id. at pg. 76-78.
[viii] Id. at pg. 74.
[ix] Id. at pg. 75.
[x] Dana-Farber Cancer Inst., Inc. v. Ono Pharm. Co., 379 F. Supp. 3d 53 (D. Mass. 2019).