Artificial Intelligence Inventor Asks If ‘WHO’ Can Be an Inventor Is the Wrong Question?

“If an Artificial Intelligence can perform the functions required to be an inventor and is not expressly excluded by the definitions in the statutes, why couldn’t an AI be credited as an inventor?” a group out of the University of Surrey provided a new challenge to the definition of inventor, asking “who what may be an inventor on a patent?” The group has created an artificial intelligence (AI) named DABUS. Using a first system of networks to generate new ideas, and second system of networks to determine consequences, DABUS invented a beverage container and a flashing device used for search and rescue that are the subjects of patent applications filed in the United States and Europe (which have not yet been published).

The Air Bud Scenario

Regarding who can be an inventor, the America Invents Act (AIA), 35 U.S.C. § 100(f) defines the term inventor is as: “the individual or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention.” Ryan Abbott, the leader of the University of Surrey group, hoping for an Air Bud  scenario (a Disney movie about a golden retriever who is able to join a basketball team when it is discovered there is no rule expressly prohibiting dogs from playing in the league) said “In these applications, the AI has functionally fulfilled the conceptual act that forms the basis for inventorship.” The term “individual” allows for a broad interpretation that does not necessarily exclude AI from being credited as an inventor. If an AI can perform the functions required to be an inventor and is not expressly excluded by the definitions in the statutes, why couldn’t an AI be credited as an inventor?

Abbott further asserts the need for and benefit of AI in solving many of the world’s problems, such as curing diseases and reversing climate change. Abbott sees the current patent laws as an obstruction to such potential developments, stating, “[I]f outdated IP laws around the world don’t respond quickly to the rise of the inventive machine, the lack of incentive for AI developers could stand in the way of a new era of spectacular human endeavor.”

A Question for the Courts

However, would an interpretation of inventorship that excludes AI really discourage developers and hinder AI innovation, as Abbott suggests? There may be alternatives that do not require a non-traditional interpretation of the statutory language, and do not negatively impact the motivation of AI developers. For example, if developers were credited with inventorship, then they may still be motivated to develop AI. Rather, Abbott’s conclusion seems to undercut his argument, because crediting the AI instead of the developer may act as more of a deterrent for development than the other way around. If developers do not receive credit for the products produced by their work, then it stands to reason that they would be less inclined to perform their work.

In answering this question, the courts may look to other portions of the relevant statutes. Although, the definitions in 35 U.S.C. § 100 use the broad term “individual,” other portions of relevant law differ by using words like “person” and “himself or herself.” For example, 35 U.S.C. § 102 states “A person shall be entitled to a patent unless…” In this instance, the statute has used “person” instead of “individual.” Critics of AI inventorship may argue that the use of “person” throughout patent statutes provides clarification to the definition and supports an interpretation that the law intended to limit inventorship to “persons.”

Conversely, Abbott and AI inventorship apologists will likely advocate for an interpretation that emphasizes the distinction between “individual,” as used in the definitions, and “person” as used in other portions of the statutes. Apologists may argue that, similar to claim construction, two different terms have two different meanings, and if law makers had intended to limit the definition of an inventor to a person, then they would have used the term “person,” as they did in 35 U.S.C. § 102.

AI Isn’t Going Away

Although people like Elon Musk may view AI as “our biggest existential threat,” and wish that its development would stop, the number of patent applications regarding AI and machine learning has been quickly increasing (e.g., the number of mentions of deep learning in patent filings at the World Intellectual Property Organization (WIPO) grew annually at an average rate of 175% from 2013-2016). With such strong development in the fields of AI and machine learning, it is clear that DABUS will not be the only instance of AI inventorship, and the interpretation of inventorship will likely need to be resolved in the courts.



Warning & Disclaimer: The pages, articles and comments on do not constitute legal advice, nor do they create any attorney-client relationship. The articles published express the personal opinion and views of the author as of the time of publication and should not be attributed to the author’s employer, clients or the sponsors of

Join the Discussion

7 comments so far.

  • [Avatar for Yuriy Tolmachev]
    Yuriy Tolmachev
    August 13, 2019 07:29 am

    The biggest issue is Non-obviousness: if anyone can run the key-predicting software (once the structure of the lock becomes publicly known), it is likely that several patent applications
    with similar composition-of-matter claims will be filed nearly simultaneously, and the USPTO will reject all of them for obviousness
    ( ),
    for the greater benefit of generic manufacturers ( ) and of patients, who will not have to pay
    exorbitant prices for it ( ).

  • [Avatar for Yuriy Tolmachev]
    Yuriy Tolmachev
    August 13, 2019 07:13 am

    I think the question who is the inventor in this case has an unambiguous answer.
    1) the US Constitution gives the Congress the power “to grant exclisive rights .. to inventors” (
    Since only humans (at least in the USA) can have property rights, the AI will have no rights to the invention.
    2) Let’s see now if there is a human inventor with property rights in this process.
    Since “patentability shall not be negated by the manner in which the invention was made” ,
    the fact that AI was involved in the inventive process, does not matter.

    This situation is no diferent from the process that creates most modern pharmaceuticals:
    One software predicts possible molecular “keys” for a known molecular “lock”, another software predicts possible routes to make such keys.
    A machine prepares in combinatorial fashion a plurality of the molecular entitties. They are screened in a standard (non-inventive) fashion using
    available in vitro, Petri dish or animal models. Some of the molecular entities are found to have the desired activity, some are not.

    Which human becomes the inventor in this scheme depends on the claims of the patent that gets issued eventually.
    For composition-of-matter (CoM) claims, the people who ran the “key”-generating software are the legal inventors.
    The people who set up the software instructions for the synthesis machine are the inventors on methods of making and on CoM claims (the latter is due the enablement requirement).
    The people who ran the efficacy tests are NOT inventors, since they did not contribute to the conception of any claim.
    AI does not change anything in this scheme.

  • [Avatar for Yuriy Tolmachev]
    Yuriy Tolmachev
    August 13, 2019 06:25 am

    Let’ not forget 35 U.S. Code §?103. Conditions for patentability:
    Patentability shall not be negated by the manner in which the invention was made.
    From this it follows that the people who ran the inventive software, can apply for a patent in their own names. In fact, a situation when the structural formula for a medication is “invented” first thru computer simulations, is currently the most common route to invent pharmaceuticals. Another computation generates possible routes to synthesizing this medication. The human role is to make (often with the assistance of automatic synthesis machines) the proposed molecule, and to prove its efficacy in a standard in vitro, Petri dish or animal model. I do no see any problem with the question who the inventor is in this case- it is the people who ran the softwares (the other members of the team -synthetic chemists and animal technicians would not be inventors on the compositions of matter claims, unless they made contributions beyond what the sofwares predicted). My bigger concern is not with inventorship but with non-obviousness:
    if anyone can generate these compositions of matter (with a reasonable expectation of success, then the invention is obvious, and no patent can be issued.

  • [Avatar for John V. Martinez]
    John V. Martinez
    August 7, 2019 09:42 am

    Did the AI in question sign any documents assigning ownership of its inventions to somebody else? Or did the AI directly contact the lawyer and ask it to file the patent on its behalf? If neither of those things happened, is it even legal for someone else to file that patent application? 🙂

  • [Avatar for Anon]
    August 5, 2019 08:24 pm

    Sad to see an entire article penned without the realization of the (critical) difference between a real person and a juristic person.

    Did the author happen to be asleep through Stanford v. Roche?

  • [Avatar for Anon2]
    August 5, 2019 03:44 pm

    The person who programs an algorithm, and sets it into motion causes the results of the algorithm. As such she is the author and sole sentient and conscious cause of the results of the algorithm.

    Being the first sentient conscious being who recognizes that the results of the nonconscious algorithm are useful, new, and unobvious, she also “discovers” it.

    There is plenty to reason to attach “inventorship” to such a person who creates, guides, and monitors the toys of our age.

    There is perhaps some difficulty when the algorithm is authored by certain individuals who may differ from individuals who might be responsible for its training, operation, or application to a certain problem, who in turn might differ from those individuals who first “discover” some products of the algorithm constitute an invention. In such cases likely all person should be considered joint inventors.

    Perhaps one day actual artificial consciousness (not merely simulation of consciousness’s outward appearance or the generation of similars to what consciousness can produce) will pose a real problem for the patent system…in fact all laws and the structure of society… but we are very far from that day. IMHO Only when free-will, the mind-body problem, and consciousness in human brains, are all fully scientifically understood, might we be able to recreate the same in a non-biological system… and only then, if we succeed in creating one which is capable of invention,would it become plausible that that an individual non-human inventor should be considered.

  • [Avatar for MaxDrei]
    August 5, 2019 02:22 pm

    Abbott sees AI as capable of finding solutions to “many of the world’s problems”. In Europe, where Abbott is based, patentability is limited to subject matter which solves a technical problem. In the UK, where Abbott is based, the only entity eligible to be named as inventor is “the actual devisor” of the claimed subject matter. I suppose that, just as a digital computer can “devise” a chess move, so it can devise a solution to a technical problem.