Artificial intelligence is being used more and more to develop new inventions without human aid. But as AI capabilities increase at a rapid pace, patent offices around the world face a crucial question: can AI be an inventor for purposes of patent protection?
The Artificial Inventor Project (AIP) – which describes itself as a group dedicated to “seeking intellectual property rights for the autonomous output of artificial intelligence” – recently put this question to the test in various patent offices, including the U.S. Patent and Trademark Office (USPTO), the European Patent Office (EPO) and the United Kingdom Intellectual Property Office (UKIPO).
On July 29, 2019, the AIP filed patent applications naming the “Device for the Autonomous Bootstrapping of Unified Sentience” (DABUS) AI system as the sole inventor. Specifically, the DABUS application listed the inventor’s given name as “[DABUS]” and the family name as “Invention generated by artificial intelligence.” DABUS is a patented AI system created by Dr. Stephen Thaler. The DABUS application listed Dr. Thaler as the Applicant-Assignee. To meet the Oath/Declaration requirement of the USPTO, Dr. Thaler submitted a Substitute Statement in Lieu of an Oath or Declaration. Dr. Thaler also submitted a Statement under 37 CFR 3.73(c) identifying himself as the assignee of the entire right, title and interest in the application, and executed an assignment on behalf of both DABUS (as the assignor) and himself (as the assignee) to assign the entire right, title and interest in the invention to himself.
On Aug. 8, 2019, after preliminary review of the application documents, the USPTO issued a Notice to File Missing Parts indicating that the Application Data Sheet (ADS) filed with the application was deficient because it “does not identify each inventor by his or her legal name.” On Aug. 29, 2019, Dr. Thaler filed a petition under 37 CFR 1.181 requesting supervisory review of the Aug. 8 Notice and requesting that it be vacated. The USPTO issued a second Notice to File Missing Parts on Dec. 13, 2019, and on Dec. 17, 2019 dismissed the Aug. 29 petition.
On Jan. 20, 2020, Dr. Thaler filed another petition requesting reconsideration of the Dec. 17 dismissal of the Aug. 29 petition. The Jan. 20 petition called for the USPTO to consider the question: can a machine invent under U.S. patent laws? Dr. Thaler argued that DABUS is a creativity machine designed to invent, not to solve a particular problem, and because DABUS itself “recognized the novelty and salience” of the invention, DABUS (the machine) was properly identified as the inventor of the invention.
After consideration of Dr. Thaler’s Jan. 20 petition, the USPTO ultimately found that an inventor must be a natural person. To support this decision, the USPTO cited existing U.S. statutes, U.S. case law and the Manual of Patent Examining Procedure (MPEP).
The following are the statutes cited in the USPTO’s decision:
- 35 U.S.C. § l00(f) provides: The term “inventor” means the individual or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention.
- 35 U.S.C. § l00(g) provides: The terms “joint inventor” and “coinventor” mean any 1 of the individuals who invented or discovered the subject matter of a joint invention.
- 35 U.S.C. § 101 provides: Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
- 35 U.S.C. § 115(a) provides: An application for patent that is filed under section 111 (a) or commences the national stage under section 371 shall include, or be amended to include, the name of the inventor for any invention claimed in the application. Except as otherwise provided in this section, each individual who is the inventor or a joint inventor of a claimed invention in an application for patent shall execute an oath or declaration in connection with the application.
- 35 U.S.C. § 115(b) provides, in pertinent part: An oath or declaration under subsection (a) shall contain statements that … such individual believes himself or herself to be the original inventor or an original joint inventor of a claimed invention in the application.
- 35 U.S.C. § 1 l 5(h)(l) provides, in pertinent part: Any person making a statement required under this section may withdraw, replace, or otherwise correct the statement at any time.
The USPTO noted that these patent statutes repeatedly use terms such as “whoever” and “individual” and “person,” and thus Title 35 of the U.S. Code consistently refers to inventors as being natural persons. The USPTO therefore found that interpreting the patent statutes broadly to encompass machines, as well as natural persons, would be contrary to the plain text of the statutes.
The USPTO also noted that federal courts have referred to inventors as natural persons as well. In its decision, the USPTO cited Univ. of Utah v. Max-Planck-Gesellschaft zur Förderung der Wissenschaflen e.V., where the Federal Circuit held that a state could not be an inventor (734 F.3d 1315, 1323 (Fed. Cir. 2013)), and Beech Aircraft Corp. v. EDO Corp., where the Federal Circuit held that a corporation could not be an inventor (990 F. 2d 1237, 1248 (Fed. Cir. 1993)). While these Federal Circuit decisions did not address AI as an inventor, the USPTO found that the Court’s discussion of conception in the context of corporations and states also was applicable to machines, and thus supported the position that conception of an invention must be by a natural person.
The USPTO found that this position was consistent with Title 37 of the Code of Federal Regulations, which repeatedly references an inventor as a “person,” as well as the MPEP which includes numerous instances of terms such as “mental” and “mind,” thus again indicating that conception must be by a natural person.
The USPTO therefore ultimately held that an “inventor” under current U.S. patent law can be only a “natural person.”
Dr. Thaler has filed a complaint in the Eastern District of Virginia asking the court to declare that the USPTO should not reject AI-generated inventions which name AI as an inventor.
Until the USPTO, courts and/or legislature provide further guidance, the question remains as to how to handle inventions that were partly or fully developed or conceived of by AI.
In cases where both a human and AI conceived different parts of an invention, or where the invention is an implementation of AI or implemented by AI, one strategy would be to file the patent application naming both the human and AI as joint inventors. For example, applications directed to the output resulting from human interaction with the AI system or where the invention relies upon human interpretation of the output of the AI system may follow this strategy. However, this strategy should be employed only when the human has made a contribution to the invention.
Since joint inventors are co-owners of a patent under U.S. patent laws, the human inventor would be able to claim full ownership of the patent rights, regardless of the level of his or her contribution. In the case of the DABUS invention, for example, a single patent application could have been filed including claims directed to DABUS itself as well as claims directed to the inventions invented by DABUS, and listed both DABUS and a human as joint inventors.
Another strategy may be to file a patent application listing only a human as an inventor, even where AI is a partial or full inventor. It’s important to note that under U.S. patent laws, a patent which is premised on incorrect inventorship can result in the patent being held invalid or unenforceable, or even in a fine or imprisonment for an inventor who had incorrectly signed a declaration of inventorship (see e.g., Ethicon, Inc. v. U.S. Surgical, 135 F.3d 1456, 1465 (Fed. Cir. 1998)).
Therefore, this approach only should be employed as a measure to preserve rights in the invention, particularly in cases where an applicant has publicly disclosed or is about to publicly disclose the invention, until we have further clarity on AI inventorship. If an applicant follows this approach, the human inventor should delay submitting a declaration of inventorship until the allowance stage, at which point inventorship is evaluated, with the hopes that the laws surrounding AI inventorship are clarified. If it ultimately is decided that AI can be an inventor for purposes of a patent application, the applicant may then amend the inventorship to include the AI inventor and submit the declaration of inventorship. If it is decided that AI cannot be an inventor on a patent application, the applicant likely will be left with no choice but to abandon the application.
Yet another approach would be not to file a patent application and instead rely on trade secret protection for the AI-generated invention. This is a viable option only if the invention can be maintained in secrecy, meaning public disclosure can be prevented. To maintain the invention as a trade secret, knowledge of the invention should be limited to “need-to-know” employees and such employees should be subject to non-disclosure agreements. One benefit of this approach is that trade secrets potentially can have an infinite duration, assuming they are maintained in secrecy, as opposed to patents, which are limited to a 20-year term. It is important to keep in mind that while trade secret protection may be effective for certain AI inventions, such as those where only the input and the output are detectable, but would be ineffective for other AI inventions such as those that can be reverse engineered.
It is becoming more and more evident that judicial decisions and/or legislative changes will be needed to catch up with today’s technological advancements, particularly in view of the increasing number of AI-related patent applications being filed. On Oct. 27, 2020, the USPTO released a report showing that the number of AI patent applications filed in certain technology categories has increased by more than 100 percent between 2002 and 2018. Those categories are vision, planning/control, knowledge processing, speech, AI hardware, evolutionary computation, natural language processing, and machine learning. The number of AI patent applications in these categories increased from 30,000 in 2002 to more than 60,000 in 2018. Further, the Oct. 27 report showed that the overall share of AI patent applications increased from 9 percent in 2002 to nearly 16 percent in 2018.
The USPTO also released another report on Oct. 6, 2020 entitled “Public Views on Artificial Intelligence and Intellectual Property Policy,” in which it analyzed nearly 200 responses from individuals and organizations on this issue. The consensus of the responses appeared to be that, for now, only humans, not machines, can be inventors. Also, most of the responses agreed that only a natural person or a company, through an assignment, should be considered the owner of a patent.
Should they occur, legislative reforms may, for example, recognize AI systems as inventors or may permit the naming of AI as an inventor provided there is some defined interaction or relationship with a human operator. However, we have yet to see if any such reforms will be enacted. Therefore, for the time being, it might be best to consider AI as an inventing tool or aid, not an inventor itself.