By Andrew Costa, ND Galli Law
As artificial intelligence (“AI”) and machine learning become ever more prevalent, new questions continue to emerge as intellectual property law tries to keep pace with these advancing technologies. AI and machine learning are programs that mimic how the human brain processes information and then use the information collected to continually improve functionality.
AI can be simple, like Netflix’s Recommendation Engine that predicts a user’s movie preferences based on the information the program gains about the user; or they can be much more complex, like DABUS that mirrors the operation of neurons in the brain to perform a variety of tasks and invent new things. One issue that has arisen is whether the AI can be named as the inventor or co-inventor on a patent application despite nearly universal statutory language that inventorship must originate in a natural person.
The ultimate question is whether the language of patent laws around the globe can be read broadly enough to permit an AI to be an inventor, or at least whether a person who uses an AI system name himself the inventor if he employs the system but does not invent the invention himself. Thus far, with two notable exceptions (South Africa and Australia) most jurisdictions across the world agree that a natural person must still be the named inventor, and that AI systems do not have cognizable legal rights that would permit an assignment from the AI to the user. The initial success in Australia was recently reversed by the Australian Federal Court, however, which we discuss in some detail below.
What is DABUS?
One scientist, Dr. Stephen L. Thaler, is leading the push around the globe for legal recognition of AI and machine learning as inventors for the purposes of patentability. Thaler listed the AI system called DABUS, which stands for “Device for the Autonomous Bootstrapping of Unified Sentience”, as the inventor of a beverage container on a patent application with the USPTO in 2019, while naming himself the legal representative of the machine and the assignee of the application. The USPTO, in a decision affirmed by the US District Court for the Eastern District of Virginia, rejected this application for failing to name an inventor, on the grounds that inventors must be “natural persons.” The case is now on appeal again, pending before the US Court of Appeals for the Federal Circuit.
Dr. Thaler has also filed patent applications in other countries across the globe for the same beverage container, with DABUS as the sole named inventor, in an effort to persuade these jurisdictions to acknowledge AI and Machine Learning systems as inventors on patents. Although the majority of jurisdictions outside the United States have denied Dr. Thaler’s applications for failing the same statutory inventorship requirements, including in the United Kingdom, the European Patent Office and Taiwan, Dr. Thaler’s applications in South Africa and Australia were successful, at least initially, and the initial rejection in Germany was just set aside. We discuss the situation Australia and Germany in more detail below.
Australia: Victory, Failure and an Open Door
The decision of the Australian Federal Court is the most recent event in Dr. Thaler’s saga to have his DABUS AI recognized as an inventor on a patent. The Australian Deputy Commissioner of Patents originally rejected Thaler’s application, but on appeal, Judge Beach, a primary judge of the Federal Court, reversed giving Thaler an early victory. This first appeal held that DABUS could be the inventor under Australian law, and that an assignment of patent rights to Thaler was unnecessary because DABUS operates under Thaler’s ownership and control. Consequently the court reasoned, Thaler derives ownership rights in anything DABUS invents under Section 15(1)(c) of Australia’s Patent Act that permits a party acquire title to a patent without an assignment. In short, the court proffered three reasons DABUS could be the inventor: (1) the term “inventor” in the Act is an “agent noun” and thus the term should be read to include “a person or thing that invents;” (2) to prevent an AI from being an inventor would, as a matter of policy, preclude “many otherwise patentable inventions where it cannot sensibly be said that a human is the inventor;” and (3) through employing a canon of construction that teaches against a court reading a limitation into a statute unless those limitations give it purpose, “nothing in the Act dictates the contrary”. The Australian Deputy Commissioner of Patents appealed this decision to a full panel of the Federal Court.
In a unanimous panel decision by five judges of the Federal Court of Australia, the court overruled Judge Beach, definitively holding that an AI system cannot, as a matter of Australian law, be the named inventor of a patent, and nor could that AI assign invention rights to Thaler. Reasoning from historical precedent and modern Australian common law, the court explained that the Patent Act bases invention on the idea that it arises “from the mind of a natural person or persons.” Interestingly, the court cautioned using statutory construction as a means to read “desirable policy” into legislation and from that reading infer the policy as the purpose of the legislation.
The court expressed concerns that: “If an artificial intelligence is capable of being recognized as an inventor, should the standard of inventive step be recalibrated such that it is no longer judged by reference to the knowledge and thought processes of the hypothetical uninventive skilled worker in the field? If so, how? What continuing role might the ground of revocation for false suggestion or misrepresentation have, in circumstances where the inventor is a machine?” The court found these consequences untenable.
The court then addressed assignment and held that a plain reading of the Act only permits an assignment of patent rights between parties with a legal identity. The court concluded that unlike a business, for example, DABUS, as an AI system has no cognizable legal identity and therefore could not be an inventor for the purposes of a patent, nor could it effectuate an assignment.
Though this decision is consistent with those in other jurisdictions around the world, interestingly the Australian Federal Court did not decide whether Thaler could derive rights to an AI-created invention under 15(1)(c). The court clarified that determining a patent’s inventor is a question of law, and because Dr. Thaler never alleged he was the inventor, the appeals court could not address derivation. Making no decision, the court noted how Dr. Thaler is responsible for maintaining, operating and financing DABUS, and owns DABUS’s source code and its copyright. Perhaps this language insinuates an openness in the court to permit a natural person to receive a patent for the inventions of an AI if conditions like these are met.
The Uncertain Path Forward for AI Inventorship
Despite the clear setbacks in the majority of jurisdictions who have considered the issue, Dr. Thaler’s attempts to patent inventions naming DABUS as the inventor are not completely finished. As noted, Dr. Thaler is currently appealing the US loss to the United States Court of Appeals for the Federal Circuit, and in Australia he may still seek “special leave” to appeal to the Australian High Court, the equivalent of the United States Supreme Court.
In Germany, a March 2022 decision of the German Federal Patent Court set aside a ruling by the Patent and Trademark Office (DPMA) that refused Thaler a patent application for DABUS. The court explained that because German patent regulations do not overtly prohibit the inclusion of an AI in addition to a human inventor on a patent, Dr. Thaler could list himself as the inventor and at least name DABUS as a tool within the same section as the AI he used to invent the subject matter. This decision by the German Federal Patent Court may, at least, acknowledge the contributions of AI to inventorship and allow inventors who collaborate with AI to patent their inventions. This aligns with a 2020 resolution by the International Association for the Protection of Intellectual Property (AIPPI) that argues against a per se rule precluding inventions that incorporate AI-contributions from patent protection. The AIPPI explains in that resolution that AI should not be considered an inventor or co-inventor in the absence of a human contribution, but would instead permit a natural person to claim an AI-derived invention if that natural person has made an “intellectual contribution to the inventive concept.”
In most jurisdictions, however, without a profound shift in reasoning among the highest courts of these countries in favor of a broader interpretation of the requirements of the patent laws, legislative action would be required to acknowledge AI as an inventor. As these cases show, absent a dramatic shift in statutory interpretation or legislative action, currently most patent laws are seen to only protect the ingenuity of human inventions. Although AI may mimic human cognition exceedingly well, patenting is still within the exclusive domain of human beings.