A.I.Nventor

The new en banc petition in Thaler v. Vidal offers potential for future development on the law of invention and inventorship. 

by Dennis Crouch

Hi, my name is Dennis, I am a natural person, a human being, an individual.  But, I also think of myself as a collective–trillions of cells and other biologic matter, only some of which expresses “my” DNA; a host of personalities all housed within a thick skull.

In my view, it is unquestionable that AI regularly contribute to inventive concepts so substantially as to be named joint-inventors alongside their human counterparts, if it were permitted.  Many of us are hung-up on the notion that inventorship requires “conception in the mind”–a feat perhaps beyond any computer today.  But conception is not a requirement for joint inventorship. Cases like Dana-Farber v. Ono help us understand how it might work. In Dana-Farber case, some of the joint inventors (our AI equivalents) provided data and analysis, but then the actual “conception” was done by a third party after receiving the data inputs.  Even though only one of the inventors actually “conceived,” the Federal Circuit held that all three should be listed as inventors because each made substantial contributions that led to the conception. Dana-Farber Cancer Inst., Inc. v. Ono Pharm. Co., Ltd., 964 F.3d 1365 (Fed. Cir. 2020). See Toshiko Takenaka, Unravelling Inventorship, 21 Chi.-Kent J. Intell. Prop. 71 (2022).

The key legal justification for excluding AI is the traditional human-only rule of inventorship. Although the Patent Act does not expressly declare “inventors-must-be-humans” or “no-robots,” since 2011 it has stated that inventors are “individuals.”

(f) The term “inventor” means the individual or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention.

(g) 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. § 100(f)/(g) (2011).  Earlier this summer, the Federal Circuit sided against Dr. Stephen Thaler on this issue. Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022). Thaler has been attempting to seek patent protection for two inventions created by DABUS, his AI computer. Id. In the case, the USPTO admitted (for the purposes of the litigation) that DABUS had conceived of an invention.  Thus, the only question on appeal was whether the USPTO acted appropriately in denying patent protection solely based upon the fact that the purported inventor is non-human. Id.

The court sided with USPTO’s no-patent stance. It held that US Patent Laws require listing of an inventor, and that inventor must be a “natural person” — i.e., human being. Although the court cited several statutory justifications for its decision, the key factor came from the definitions found in 35 U.S.C. § 100 that identify inventors as “individuals.” The court concluded that the term “individual” is best interpreted as limited to a human being.  Thaler, 43 F.4th  at 1211.  One quirk of this ruling is that this ‘individual’ definition was added quite recently as part of the 2011 America Invents Act and nothing in the legislative history indicates an intent to use the term to exclude AI from inventorship rolls.

Through his attorney (Professor Ryan Abbott), Thaler has now petitioned the Federal Circuit for en banc rehearing on the following question:

Whether an artificial intelligence can be an inventor for purposes of patent law, which implicates the most fundamental aspects of patent law, namely, the nature of inventorship and therefore whether AI discoveries can be patented.

Thaler en banc petition. The petition makes three key complaints against the petition, which I paraphrase below:

  1. The panel selectively quoted from dictionaries for its conclusion that individuals are always human.  The better definition of individual is a “distinct, indivisible entity”–a definition that would include an AI-inventor.  I might ask, if an AI is not an individual, does that mean it is a collective?
  2. The panel unduly disregard of the Patent Act’s promise of patent rights regardless of “the manner in which the invention was made” and without limiting eligibility scope only to areas contemplated by Congress. See 35 U.S.C. 103 (“Patentability shall not be negated by the manner in which the invention was made”); 35 U.S.C. 101 (“Whoever invents or discovers any new and useful process …”); Diamond v. Chakrabarty, 447 U. S. 303 (1980) (that the patent system allows for “inventions in areas not contemplated by Congress.”).
  3. In the face of evolving technology that redefines the potential of inventorship, the panel failed to interpret the statute “in light of [the] basic purpose” of the Patent Act.  Quoting Twentieth Century Music Corp. v. Aiken, 422 U.S. 151 (1975).

Id.  In some ways, Judge Stark’s opinion reads like a district court applying the law handed-down rather than an appellate court with a larger role of considering meaning and importance of ongoing precedent. Perhaps this makes sense, Thaler was the first precedential opinion authored by Judge Stark’s since joining the Court of Appeals earlier in 2022.  Judge Stark was previously a district court judge in Delaware for more than a decade.

I have always seen the final sentence of Section 103 as an important declaration of patent law policy that goes beyond simply obviousness doctrine. “Patentability shall not be negated by the manner in which the invention was made.” One particular bit of the Federal Circuit decision that gets in my craw is the court’s statement limiting that provision as only applicable in the obviousness context. In my view, the statement should be applicable in the Thaler analysis as well as in other contexts, such as patent eligibility. I’m planning a subsequent post that will focus more on this issue.

If my calculations are correct, any Amicus Brief in support of Thaler (or Support of Neither Party) would need to be filed by Oct 3 (absent extension). 

The absence of a dissent in the original opinion makes the odds of overturning the panel decision quite low–Thaler would need seven of the remaining nine active judges (assuming that none of the original three change their minds).  But, Thaler may still appreciate an outcome where one or two judges offer some additional commentary that would then serve as fodder for the upcoming petition for writ of certiorari.

70 thoughts on “A.I.Nventor

  1. 18

    Setting aside for a moment the “slave” question, what about inventions made outside the jurisdiction. I am old enough to remember the days of “local novelty” and the so-called “communication invention”. That was when a UK patent attorney received from outside the jurisdiction a “communication” that included an enabling disclosure of a patentable invention. Under the law at the time, the attorney would file in their own name, naming themself as “inventor”. This was considered OK, and in harmony with the aims of the patent statute, to promote the progress of the useful arts, within the jurisdiction.

    By analogy, when an AI communicates a patentable invention to a legal person, it can be seen as legitimate to name that legal person as the inventor.

    1. 18.1

      Certainly our statute could be amended to work like that. Our 1952 act distinguished between inventions invented in the U.S. and outside the U.S. when assessing novelty, so the idea has analogies in our own legal history. There is scant sound basis—either in the statutory text itself or the subsequent jurisprudence—that the statute currently works like that.

      1. 18.1.1

        Greg,

        The question of Being an inventor was not affected by the “prior art must be US-based” mechanism you share.

        You are conflating things rather than clarifying things.

      2. 18.1.2

        Indeed. The concept of the “communication invention” is so far in the distant past by now that Googling it produced not one single relevant hit (amongst a mass of irrelevant hits). It seems that,these days, nobody but me has ever heard of it. I mentioned it only as a possibility for amending the Statute to manage the burgeoning field of AI-invented subject matter, each time it emerges out of an alien (cyber) world and enters the world of human perception.

        1. 18.1.2.1

          That just won’t square with how the law is.

          This has been made clear to you with my “black box in a second room” example and its derivations.

          Pretending otherwise is simply not helpful.

    2. 18.2

      Not under current US law, MaxDrei — as you well know by now (or at the least, Should well know by now).

      The variant of my “second person opens a black box into which an invention of another has been placed” scorches your “attempt” here.

      That variant being rather than a single person “opening the box,” the opening is live-streamed over the internet to ‘millions.’

      US law simply does not let “mere reading” rise to the level of meeting the legal requirement to be an inventor.

      It is beyond curious as to why such canards are still bandied about.

      1. 18.2.1

        As I just explained to Greg, I was thinking how a patent statute might be tweaked to render it capable of handling inventions made by an AI. Is there any patent statute currently in force anywhere in the world that accommodates inventions made by an AI?

        1. 18.2.1.1

          So far (as I recall), only South Africa with its semi-registration system has been the only “success” for DABUS.

          And I need not remind you of the UK’s acceptance of DABUS to a certain (non-patent grant) state, eh?

      1. 16.1.1

        Do we need a guardian ad litem for the AI? Is the AI old enough to sign the oath? Are there child labor issues here?

        Also, if the AI itself is prior art, then does that make the invention obvious (assuming there were no co-inventors maybe)? Was it obvious to try, i.e., run the AI?

        1. 16.1.1.1

          How do you punish the AI for false statements wrt the oath (“…willful false statements and the like are punishable by fine or imprisonment, or both…”)?

          The death penalty, i.e., only punishment that seems relevant to an AI??

  2. 15

    I’m not sure I understand. Just because a machine contributes and accelerates the process for inventing something doesn’t mean the machine actually “creates a novel idea.” A human being still needs to look, review and analyze the data. The machine never has a “spark of genius” like human have. “Conception is the touchstone of inventorship, the completion of the mental part of the invention.” Burroughs Wellcome Co. v. Barr Labs., Inc., 40 F.3d 1223

    1. 15.1

      There is no “spark of genius” in dumping data into teaching a learning network nor separating the chaff from the wheat in the AI’s output. And yes Congress expanding inventor in 35 USC 100 to “individual entity” would encompass “things” including AIs, opening all sorts of can’s of worms from excluding natural persons.

  3. 14

    “I have always seen the final sentence of Section 103 as an important declaration of patent law policy that goes beyond simply obviousness doctrine. ”

    Common misconception, you really should stop spreading that around.

    “One particular bit of the Federal Circuit decision that gets in my craw is the court’s statement limiting that provision as only applicable in the obviousness context.”

    Based.

    Obviously if they wanted it to be a general provision it should have gone elsewhere in the statutes.

    1. 14.1

      6 – up to your old silly tricks?

      Go back and review the (many) discussions on the Act of 1952, the Choice to not use “invention” and INSTEAD use obviousness.

      As a helpful reminder, what is NOW the multiple sections of (at least) 101 and 104 was but a single same paragraph previously.

      1. 14.1.2

        “As a helpful reminder, what is NOW the multiple sections of (at least) 101 and 104 was but a single same paragraph previously.”

        Then they should have formatted and/or written it better during the split. Derp. Statute drafting isn’t rocket science.

        1. 14.1.2.1

          Your “then they should have” does NOT compensate for Your error in accusing Crouch of perpetuating a misconception.

          You’ve been around long enough that you should know better.

          Instead, your accusation is merely a projection of your own misinformation.

          Not a good look, 6.

  4. 13

    Dennis,

    “Congress shall have power … To promote the Progress of Science and useful
    Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries;”

    If a case like Thaler’s were to be reviewed and decided by SCOTUS, which – if any – of the various possible modes of interpretation (textualism, original meaning, judicial precedent, pragmatism, moral reasoning, national ethos, structuralism, historical practices, etc.) would actually permit “Authors and Inventors” to be construed as including AI? With the possible exception of Franklin and Jefferson, the founders would not have seen it that way. Given the interpretive approaches favored by our current justices, I don’t see it as having any likelihood of success. If it were to be taken up, it might even result in a 9-0 decision.

    1. 13.1

      Mark – I’ll just note here that for 100+ years our copyright law has allowed for non-humans to be identified as the author as part of our work-made-for-hire doctrine.

      1. 13.1.2

        Let me backtrack slightly. From the Compendium of US Copyright Office Practices (3rd ed., 2021), 306 and 313.2:

        The U.S. Copyright Office will register an original work of authorship, provided that the work was created by a human being. The copyright law only protects “the fruits of intellectual labor” that “are founded in the creative powers of the mind.” Trade-Mark Cases, 100 U.S. 82, 94 (1879). Because copyright law is limited to “original intellectual conceptions of the author,” the Office will refuse to register a claim if it determines that a human being did not create the work. Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53, 58 (1884).

        The crucial question is “whether the ‘work’ is basically one of human authorship, with the computer [or other device] merely being an assisting instrument, or whether the traditional elements of authorship in the work (literary, artistic, or musical expression or elements of selection, arrangement, etc.) were actually conceived and executed not by man but by a machine.” U.S. Copyright Office, Report to the Librarian of Congress by the Register of Copyrights 5 (1966).

  5. 12

    [I]t is unquestionable that AI regularly contribute to inventive concepts so substantially as to be named joint-inventors alongside their human counterparts, if it were permitted.

    I certainly concede that it is possible that AI is presently making such contributions, but “unquestionable” seems too strong an assertion. How would we know?

  6. 11

    “nothing in the legislative history indicates an intent to use the term to exclude AI from inventorship rolls”

    Good grief. I really want to believe this a joke but I suspect otherwise.

    1. 11.1

      Agreed; one might just as well consider “inventorship” as a participation trophy in step with the rest of the current American mindset.

      MHO, YMMV, etc. Be well.

  7. 10

    The part that struck me in the post was Dennis’s definition of “individual” as “‘distinct, indivisible entity’–a definition that would include an AI-inventor.”

    That’s begging the question IMO. It’s the “better” definition for including an “AI-inventor” because it includes an AI-inventor.

    An instantiation of AI is a tool, rather than an individual, because it is replicable. Another copy of the same software with the same training would provide the same capability to others as did the “AI-inventor” to its human “co-inventors.” Just as a second hammer would provide the same capability as a first hammer.

    We should be brave enough to interpret the patent statutes to require that the inventors be human, even in these strange times.

  8. 9

    The whole “patentability shall not be negated” thing is a red herring. The issue here has never been one of patentability, which is a question of the subject matter to be patented. It’s one of inventorship. That question may result in a de facto result that nobody (or Thaler, maybe) can actually apply for a patent on the invention, but it goes far beyond a reasonable reading of 103 to think that it applies to this situation.

    1. 9.1

      Exactly right. Under current law, an AI inventor is in the same position as a North Korean citizen—it can conceive an invention that satisfies U.S. standards of patentability, but it still cannot apply for a U.S. patent. The last sentence of §103 no more compels the USPTO to list DABUS as an inventor than it would compel the USPTO to list Kim Jong-Il.

  9. 8

    Dennis, there are many who might question your initial premise that you are a natural person and human being. Many might think you are a cyborg that runs 24/7 using input from these posts as fuel.

        1. 8.1.1.1

          Apotu,

          Your comment here – inclusive of the non-human legal fiction of the juristic person of a corporation may be closer to the mark than you may have realized.

  10. 7

    It just occurred to me that the fact that anyone has standing to litigate this case answers the question.

      1. 7.1.1

        we have a big question about who owns what

        The whole discussion turns on ownership. Somebody owns every AI. That somebody is the entity entitled to the ownership of any intellectual property that the AI may, in fact, generate.

        Deciding who invented something for the purposes of recognition is a far different exercise than deciding who should get paid for the use of an invention- or who may decide to exclude others from making or using it.

        1. 7.1.1.1

          Another comment that is just not even close:

          The whole discussion turns on ownership

          But then again, this is merely marty.

            1. 7.1.1.1.1.1

              “Beside the point”….

              Point of LAW – absolutely.

              But you be you and simply don’t bother to understand the terrain upon which you want to do battle.

  11. 6

    Almost all commentators dive into the legal arguments without significant investigation into the underlying facts. Reference to the Thaler specifications shows that they have descriptions and drawings closely resembling those of hundreds of other specifications for which human inventors are attributed. There is no description of the artificial intelligence involved, how it has been set up to make the relevant inventions and the process steps involved, and there are no flow charts showing the AI inventive operation. There is no evidence from the patent specifications themselves that AI was involved in any significant way, and strong arguments that the specifications are deficient under 35 USC 112. These points need consideration before anyone becomes upset about AI eligibility as an inventor.

    1. 6.1

      There is no description of the artificial intelligence involved, how it has been set up to make the relevant inventions and the process steps involved, and there are no flow charts showing the AI inventive operation. There is no evidence from the patent specifications themselves that AI was involved in any significant way…

      Why would one expect to find any of these details in the spec? Do you spend much time describing the inventor’s long mustache, or his high pitched laugh, or his endearing bonhomie? Do you typically include electroencephalogram readings of her brain waves, captured during the process of inventing the present invention? Why should one include these details in the spec of an AI-conceived disclosure when one never includes them for a human-conceived disclosure.

  12. 5

    Dennis you assert that:

    “……it is unquestionable that AI regularly contribute to inventive concepts so substantially as to be named joint-inventors…..”

    I blinked at that. I suppose that I have failed to take proper account of what “contribute” means in this context, under US law. I suppose you used the term “unquestionable” deliberately, to foment debate. It worked.

    It seems to me that human inventors in, say, classic pharmaceutical cases, routinely rely on tools that generate and output the data which enable human users of those tools to conceive of, and discover (express in words) concepts that amount to patentable inventions. The tools render a “contribution” that is decisive to the issue whether an invention is revealed, sure. But I don’t see such “contributions” as enough to render them joint inventors.

    How is it different with a tool sophisticated enough to warrant the appellation “AI”. What is “intelligence” in this context? If a computer is programmed to provide canned laughter for a comedy broadcast but has no idea why anything it hears is funny, and sometimes laughs in the wrong places, does it really possess any”intelligence” at all? When it suggests a molecule for further research, is that a “contribution” to a pharmaceutical invention? Is that the moment when patentable matter is conceived?

    And if the AI does only what it is instructed to do, by its human operator, does it really make the machine autonomous, enough to entitle it to be dubbed an “individual”?

    Unquestionable? I would question that. But hey, I’m not in the AI building industry so in this I would defer to those who are. You too, I guess.

    PS: Can we keep clear of the laws of ownership of patentable subject matter? They vary from jurisdiction to jurisdiction but are I think not relevant to the universal issue whether (for the time being at least) a machine can invent a concept defined by a patent claim.

    1. 5.2

      You are assuming all of your conclusions here, Max. Yes, if the machine supplies a laugh without understanding why something is funny, then it makes no sense to apply the word “intelligent” to that machine. What about the machine that can—to some rough approximation—distinguish the “funny” from the “not funny”? That is the more interesting—and relevant—example to contemplate.

      1. 5.2.1

        “That is the more interesting—and relevant—example to contemplate.”

        It’s only more relevant to the discussion you’re interested in having. Max’s machine is far more relevant than yours to the technology at hand.

    2. 5.3

      MaxDrei, I would posit that “can” is not asking the right question.

      “Can” reflecting abject capability — divorced from legal hurdles which is more apt to be described with the word “may.”

      It is easy to see this with my prior “black box” scenario.

  13. 4

    I disagree re the final sentence of Section 103. It application should remain limited to its original intent (to abrogate the “flash of creative genius” test) and context (nonobviousness).

    Relatedly, this issue should not be addressed via the courts, in my view. Only Congress should address it and with a new or amended provision of the statute tailored to it specifically.

  14. 3

    If AI is the inventor, who owns the patent? Can AI assign it? If I recall correctly (and I may not), slaves could patent their inventions but the owner of the slave did not own the patent and could not be listed as an inventor.

    1. 3.1

      You are misremembering. Slaves could not apply for patents (they were not U.S. citizens, and U.S. citizenship was a prerequisite for applications back then), nor could their owners apply on behalf of the slaves. The question of who owned such patents never arose, because such patents were not legally possible in the first place.

        1. 3.1.1.1

          Egads, blame where blame is due:

          This is an egregious misstatement by me. What was I thinking?

          Perhaps: it is well recognized that slaves could and did invent.

          Further (if I may), slaves (and freed slaves) were – in fact – fully human (as opposed to any legal standing), and with such, the Lockean foundation of US patent law still (and always) applied.

      1. 3.1.2

        Are you not, perhaps, referring to U.S. patent law as it existed once Congress got around to adjusting matters after the Civil War had commenced?

        1. 3.1.2.1

          Sorry. I do not know which Congressional action you mean. Do you have a cite to the statutes at large?

          1. 3.1.2.1.1

            Since the Union had prohibited residents of seceding states from obtaining U.S. patents, Confederate inventors had little incentive to disclose their inventions and secure a patent under these terms.Jun 1, 2012

            The Story of the Confederate States Patent Office and Its …

            It has been my understanding for a long time that “the Union had prohibited residents of seceding states from obtaining U.S. patents.”

            Don’t know where I got that idea, but apparently it’s “around.” That’s all I’ve been able to find to justify the remarks I made. The book that is referenced above is not available to me, and I’m don’t want to buy a $55 copy.

          2. 3.1.2.1.2

            The same sort of situation arises in all wartime. Can enemy aliens file for and obtain US patents? Obviously the answer is “no.” An argument that enemy aliens would be unlikely to attempt to file for US patent protection of their inventions, though true, is unpersuasive. But how that restriction is imposed is difficult to discover, by Presidential Executive Order perhaps.

    2. 3.2

      Prof. Frye wrote an interesting law review article a few years back (perhaps it is to this that you refer when you speak of “recall[ing] correctly”) about the history of slave inventors. As he explains “[t]he slave owner could not swear to be the inventor, and the slave could not take an oath at all.” In fact, the USPTO “also denied at least one patent application filed by a free African-American inventor, because African-Americans could not be citizens of the United States under Dred Scott.”

  15. 2

    In my view, it is unquestionable that AI regularly contribute to inventive concepts so substantially as to be named joint-inventors alongside their human counterparts, if it were permitted.
    Of course it is questionable. I’ve written multiple AI patent applications. I’ve prosecuted many. I have several AI patent applications sitting on my desk to write. I am very well-versed in what AI is or isn’t capable of. At this point in time, what people call “AI” is little more than a glorified electron-microscope — allowing scientists/engineers/inventors insights into phenomena that they wouldn’t otherwise have. However, we don’t see people clamoring to have electron-microscopes named as co-inventors. It would be silly on its face.

    Like an electron-microscope, what passes as AI today is but a tool that is wielded by man. When AI does become close to human intelligence, people will look back today and consider the AI today as being close to a human intelligence as a strawman is a reasonable facsimile of an operational human body. However, that will be many, many decades away.

    “Patentability shall not be negated by the manner in which the invention was made.” One particular bit of the Federal Circuit decision that gets in my craw is the court’s statement limiting that provision as only applicable in the obviousness context.
    Why? Patentability and Inventorship are different concepts. What can be patented is very different than who can be an inventor.

    But, Thaler may still appreciate an outcome where one or two judges offer some additional commentary that would then serve as fodder for the upcoming petition for writ of certiorari.
    Not a chance. Why would they subject themselves to extreme ridicule?

    The remaining portion of this comment represents my opinion. Let me start by saying that I think Thaler’s inventions are garbage. I challenge anyone who supports Thaler’s opinion to actually read the specifications of the fractal container and the neural flame. The fractal container application is a half-baked idea that isn’t enabled. The person writing the application knows little about fractals, about containers, and about manufacturing. The neural flame is much worse. This is a passage from the detailed description:

    Moreover, in a theory of how cosmic consciousness may form from inorganic matter and energy (Thaler, 1997a, 2010, 2017), the same attentional beacons may be at work between different regions of spacetime. Thus, neuron-like, flashing elements may be used as philosophical, spiritual, or religious symbols, especially when mounted atop candle- or torch-like fixtures, celebrating what may be considered deified cosmic consciousness. Such a light source may also serve as a beacon to that very cosmic consciousness most likely operating via the same neuronal signaling mechanism.
    Language such as this would make L. Ron Hubbard blush. The neural flame applications cites 11 “LITERATURE REFERENCES” – 8 of which were written by Thaler. Doesn’t it seem coincidental that DABUS created an invention in which Thaler was so extensively referenced yet Thaler wasn’t a named inventor. How does this not raise red flags to anyone defending Thaler?

    IMHO, it is plain as day that both of these inventions were conceived by Thaler. Additionally, based upon my knowledge of the capabilities of AI, I struggle to find ANYTHING in these applications that could be supplied by AI. The neural flame applications reads like Thaler’s take on Dianetics.

    Read enough of Thaler’s writings and it becomes apparent that Thaler is using this as a stunt in order to legally imbue AI with some human rights. No court in their right minds are going to touch this with a 10 foot pole. With a conservative super majority, this issue is guaranteed to be DOA at the Supreme Court. There is absolutely 0% chance that they are going to issue a ruling that imbues a computer with human characteristics.

  16. 1

    I do hope that your future work includes a nod to the legal foundation of rights being protected under the Lockean concepts.

Comments are closed.