Can a person use artificial intelligence (“AI”) as a tool to assist in generating an invention? Of course. However, the material that AI can generate is so vast, and draws on so many resources that are outside any one person’s knowledge, that it may be possible to confuse whether the person is actually engaging in “inventive” activity. Accordingly, on February 13, 2024, the US Patent and Trademark Office (USPTO) issued guidance clarifying the minimal degree of contribution for a person to be considered an inventor, when using AI as a tool.
“Artificial Intelligence,” as the term is broadly understood, has been part of patents for some time. In particular, according to WIPO, the rise of machine learning in the mid-2010s caused an explosion in AI-based patents. From 2013 to 2016, patents mentioning “machine learning” increased by 28% year over year. Mentions of “deep learning” in patent filings grew during the same time period by 175%, and mentions of neural networks grew at an average rate of 46%. During that time period, artificial intelligence was conceived of as a tool for the inventor, no different than any other computer-implemented process.
As the computing capabilities of machine learning grew, some came to believe that AI is capable of inventing independently. In 2019, Dr. Stephen Thaler filed two patent applications that were purportedly invented by an artificial intelligence system that he created, which he called DABUS. Dr. Thaler listed DABUS, rather than himself, as the inventor. Dr. Thaler claimed that he was the owner of the patent applications but not the inventor. In a series of rulings from 2020 to 2023, jurisdictions around the world (including US, EP, AU, IL, and most recently the UK) rejected Dr. Thaler’s legal contention that an AI could be an inventor under the relevant patent statutes.
While rejecting the ability of a machine to serve as the “inventor,” some of the jurisdictions speculated that Dr. Thaler himself could have qualified as the “inventor.” For example, the European Board of Appeal stated that the “Board is not aware of any case law which would prevent the user or owner of a device involved in an inventive activity to designate himself as inventor under European patent law.” (Decision J0008/20, section 4.6.6, December 21, 2021). Likewise, the Commissioner of the Israeli Patent Office concluded his decision with the following caveat: “I will clarify that this decision does not deal with the question of what human involvement is required in order for a person to be considered as an inventor in an invention made with the aid of a machine. This question will be left to the appropriate case.” (Decision of March 15, 2023, paragraph 78).
It is precisely this question that the USPTO sought to answer. Drawing on decades of precedent from inventorship disputes, the USPTO explained what activity qualifies as “inventive” when using AI as a tool, for the purposes of registering a contributor as an inventor.
In order to be listed as an inventor of an AI-assisted invention, the USPTO explained, each named individual must have made a “significant contribution” to the claimed invention. A “significant contribution” is evaluated based on the test that was historically applied to resolve inventorship disputes. This test requires that each inventor must: “(1) contribute in some significant manner to the conception … of the invention;” “(2) make a contribution to the claimed invention that is not insignificant in quality, when that contribution is measured against the dimension of the full invention,” and (3) do more than merely explain to the real inventors well-known concepts and the current state of the art.” In sum, each named inventor must have significantly contributed to “the definite and permanent idea of the complete and operative invention as it is thereafter applied in practice.” The USPTO reasoned that these factors, which were historically used to apportion inventorship between team members, apply equally well to apportioning inventorship between humans and machines.
In view of these principles, the USPTO issued a non-exhaustive list of “principles” that help inform the application of these factors to AI-assisted inventions:
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.”
2. “Merely recognizing a problem or having a general goal or research plan to pursue does not rise to the level of conception.” Thus, merely feeding a prompt to an AI system may not rise to the level of inventorship. However, “a significant contribution could be shown by the way the person constructs the prompt in view of a specific problem to elicit a particular solution from the AI system.”
3. Merely reducing an invention to practice by applying the output of an AI inquiry is not sufficient to be considered inventive. However, “a person who takes the output of an AI system and makes a significant contribution to the output to create an invention may be a proper inventor. Alternatively, a person who conducts a successful experiment using the AI system’s output could demonstrate that the person provided a significant contribution.”
4. A person who “designs, builds, or trains an AI system in view of a specific problem to elicit a particular solution could be an inventor, where the designing, building, or training of the AI system is a significant contribution to the invention created with the AI system.”
5. Simply “owning or overseeing an AI system that is used in the creation of an invention, without providing a significant contribution to the conception of the invention, does not make that person an inventor.”
These principles accord with established practice, and provide helpful guidance to the patent community.
A word of caution: everything stated above is applicable for proprietary artificial intelligence programs, which are developed and owned by the inventor or his company. Under no circumstances should an inventor share his inventive concepts with an external artificial intelligence program, including a chat bot or natural language processing (NLP) tool. Such disclosures are not considered confidential. These tools log every conversation, and use the conversations as training data for improving their natural language processing. Accordingly, such disclosures could be delivered, indirectly, to members of the public.
Finally, the most extreme test case for the USPTO’s guidance – an invention which would meet the criteria of novelty and inventive step, but lacks sufficient human input to have even a single human “inventor” – has not yet been considered by any patent office. This test case appears to be still far off. Most likely, even in cases with significant machine contribution, application of this guidance would reveal at least one human inventor. The scenario more relevant for applying the USPTO’s guidance is in selecting which team members have contributed sufficiently to the development of an invention to be considered “inventors.” Patent attorneys have been counseling their clients about these matters since the very beginning of the patent system. In this regard, the USPTO’s guidance is merely a new application of well-established principles – “old wine in new flasks.”