by Dennis Crouch
This article explores the impact of Generative AI on prior art and potential revisions to patent examination standards to address the rising tidal wave of AI-generated, often speculative, disclosures that could undermine the patent system’s integrity.
The core task of patent examination is identifying quality prior art. References must be sufficiently accessible, clear, and enabling to serve as legitimate evidence of what was previously known. Although documents are widely available today via our vast network of digital communications, there is also increasing junk in the system — documents making unsubstantiated claims that are effectively science fiction. Patent offices prefer patent documents as prior art because they are drafted to meet the strict enablement standards and filed with sworn veracity statements. Issued patents take this a step further with their imprimatur of issuance via successful examination. Many of us learned a mantra that “a prior art reference is only good for what it discloses” — but in our expanding world of deep fakes, intentional and otherwise, is face value still worth much?
In a new request for comments (RFC), the USPTO has asked the public to weigh in on these issues — particularly focusing on the impact of generative artificial intelligence (GenAI) on prior art. Services like All Prior Art are using AI to churn out and ‘publish’ many millions of generated texts, hoping some will preempt future patent applications. See my 2014 post. These disclosures are often obscure, ambiguous and technically deficient and do nothing to promote the progress of the useful arts. Still, seemingly qualify as prior art under 35 U.S.C. 102, and are presumed to be enabling.
One key trick in prior art analysis is the presumption that the reference is operable and self-enabling. In other words, the USPTO and courts generally assume that a document which qualifies as prior art contains sufficient detail to enable a person skilled in the art to practice the subject matter disclosed. You might consider a technical paper from a conference which sketches out a conceptual gearbox design (but omits specific gear ratios and material specifications). Despite the technical gaps, the presumption of enablement means the patent examiner treats the document as if it teaches a skilled engineer how to make and use the described gearbox even though it does not provide sufficient guidance. This presumption facilitates the examination process by shifting the burden to the patent applicant to prove otherwise if they contest the prior art’s completeness or applicability. It also pushes applicants to ensure that their claims include limitations not found in the references — something that is becoming increasingly difficult as the scope of prior art disclosures continues to increase.
The presumption that prior art is enabled has developed incrementally. It began with a notion that issued patents had been examined and therefore the claimed subject matter was properly enabled; then expanded to included unclaimed material in issued patents. Amgen Inc. v. Hoechst Marion Roussel, Inc., 314 F.3d 1313, 1354 (Fed.Cir.2003). Later, the Federal Circuit extended the presumption to all publications. In re Antor Media Corp., 689 F.3d 1282, 1287 (Fed. Cir. 2012). Although the presumption can be overcome with “persuasive evidence” that the reference is not enabling, it continues to be difficult to prove a negative. Still, I would like to see some empirical evidence that helps me understand the severity of this issue.
In its Request for Comments, the USPTO poses a series of fifteen questions to stakeholders, including;
- Whether AI-generated disclosures qualify as “prior art” under 35 U.S.C. § 102, and whether such treatment should depend on the degree of human involvement or curation.
- How to handle the potentially enormous volume of AI-generated prior art, and its impact on patent examination.
- Whether the presumption of enablement for prior art is warranted for AI disclosures.
- How AI prior art affects the assessment of obviousness under 35 U.S.C. § 103 and the analysis of a “person having ordinary skill in the art.”
- What new USPTO examination guidance or statutory changes may be needed.
In my mind, the fact that an AI generated the prior art is not a worry, so long as the work advanced the art. Rather, my real concern is that the outflow of publications – many of which are senseless – will gum up the patent system in nefarious ways.
1. Abundance of unread AI-generated data: Generative AI systems are generating vast amounts of synthetic data that is being “published” in the sense of being made publicly available, but much of this data may never be read or reviewed by humans. For example, I estimate that over the past 2 years AI systems have spit-out more text content than the total of all prior human-written publications. Although most of this content will never be human-read, AI-generated publications will be accessible to other AI systems – that can then incorporate the synthetic learning.
2. Fictional nature of AI-generated content: Many AI systems, especially large language models, are prone to generating content that is essentially “science fiction” – i.e., plausible-sounding but factually incorrect or divorced from reality. These typically include major gaps in reasoning or explanation that divorce the disclosure from human-shared reality. This tendency towards hallucination raises doubts in my mind about the reliability and enablement of AI-generated disclosures.
3. Motivation to Combine – an AI Strength: One of AI’s strengths is its ability to identify patterns and connections across diverse domains of information. This could potentially expand the universe of “analogous art” for purposes of obviousness analysis under 35 U.S.C. 103, as well as making it easier to find motivations to combine prior art references. At the same time, the connections made by AI may sometimes be spurious or non-sensical to human experts.
These issues make me think some about the enablement and written description requirement generally. It is improper for a patentee to claim a genus based upon a disclosure that includes a large number of inoperative species. That claim requires too much follow-on research work and so does not sufficiently disclose the invention. It seems we have a parallel situation here with AI prior art – both on the individual as well as the collective level. For the collective, AI are creating many worthless disclosures, but by creating billions of disclosures they are bound to hit upon the good ones as well. Still, those will likely be unrecognized for their worth until some later date when a human truly invents but is blocked from patenting. The difficulty lies in distinguishing between legitimate insights and spurious connections.
Ben Hattenbach & Joshua Glucoft have an interesting 2015 article on point. Patents in an Era of Infinite Monkeys and Artificial Intelligence, 19 Stan. Tech. L. Rev. 32, 42 (2015). The authors author some agreement with my analysis above — focusing on claim language they argued: “if a computer published millions of variations of claims such that all but a few were useless from a technical or grammatical perspective, then it would be easier to justify not requiring inventors to account for that sea of information.” On the other hand, “if a computer generated a focused set of high-quality variations on claim language, then it would be easier to justify folding such knowledge into the scope of the prior art.” The suggestion then is some sort of balancing test that focuses both on quality and accessibility. This is different from our current approach that is much more of an on-off switch. The authors also caution against automatically extending the presumption of enablement to all AI-generated disclosures. The presumption of enablement exists for traditional publications because we assume the authors intend to fully disclose a working invention. But with our current level of AI-generated content, especially that churned out at huge scale with minimal human curation, that assumption likely does not hold.
In his student note, Lucas Yordy focuses on some of the same issues – and argues that AI generated disclosures may decrease the patent incentive to research and disclose. The Library of Babel for Prior Art: Using Artificial Intelligence to Mass Produce Prior Art in Patent Law, 74 Vand. L. Rev. 521 (2021). But, Yordy notes the problem identified in the RFC, current patent law doctrines are ill-equipped to prevent AI-generated disclosures from rendering deserving inventions unpatentable. Like others, he calls out the enablement requirement as problematic, but he also goes on to propose a “conception” requirement for prior art to “ensure that AI-generated disclosures have actually contributed to public knowledge and have undergone some evaluation before they can render an invention unpatentable.”
In 2022, Lidiya Mishchenko published an article that went even further — arguing that even unexamined patent applications are causing prior art problems: they “occupy the patent idea space and can lead to examination [errors] and third-party search errors,” thus “contribut[ing] to costly unpredictability in the patent system more broadly by preventing others from getting a patent and by creating a temporary cloud of uncertainty.” Lidiya Mishchenko, Thank You for Not Publishing (Unexamined Patent Applications), 47 B.Y.U. L. Rev. 1563, 1564 (2022); See also, Michael McLaughlin, Computer-Generated Inventions, 101 J. Pat. & Trademark Off. Socy. 224, 239 (2019).
The USPTO is accepting comments in response to the RFC until September 29, 2024 via regulations.gov.
What are your thoughts, should Generative AI cause us to rethink prior art?
Anyone else here want to use science fiction as a factual basis for rejecting a patent,
Anyone?
Bueller?
What I want to use as a factual basis for proving a claim to be invalid is everything that falls within the state of the art, including everything already “made available” to the public; and SF books are within that state of the art.
But who cares? Has there ever been a case in real life where a USPTO Examiner (let alone a petitioner for revocation) has cited a work of science fiction against a claim, let alone prevailed with such an attack on validity?
The notional POSA doesn’t have works of SF as part of their common general knowledge and are not going to go looking into the myriad of SF published works ewith a view to finding a solution to their problem in technology, are they?
The validity attack you envisage ain’t going to happen, is it? And if so, it ain’t going to succeed, is it?
You remind me of that joke about engineers in France, when they say “OK, Ok, it doesn’t work in practice. But never mind that; in theory it ought to work”.
But come on, tell me where I’m seeing it wrong.
Where you are seeing it wrong is quite simple:
A work of fiction (science or otherwise) is not proper to use as a basis for rejecting anything in patent law.
Certainly, works of fiction (science or otherwise) may have elements that are factual.
Any such elements though — being factual — will have proper non-fiction sources.
“But who cares? Has there ever been a case in real life where a USPTO Examiner (let alone a petitioner for revocation) has cited a work of science fiction against a claim,…”
Two things:
1) obviously you cared enough to respond, and
2) I personally dealt with an examiner trying to use a source of science fiction in a 103 rejection.
(He withdrew it upon my pointing out that such was not proper).
Where is your authority for the statement that:
“A work of fiction (science or otherwise) is not proper to use as a basis for rejecting anything in patent law.”
For you it might be so obvious that no authority is needed, but not for me.
What if Hall’s initial claim was for (say) a bed comprising a mattress full of water and Heinlein in his published book proposed a bed with a mattress full of water. There are no problems to put the Heinlein proposal into effect. Accordingly, the claim is flat out anticipated. End of story.
The ‘authority’ is common sense.
Fiction – by definition – is not real.
I fully admit to some items within a work of fiction to be non-fiction. But if such ARE so, then the simple and direct path is to obtain valid non-fiction sources for establishing non-fiction elements.
Your attempt to portray my view as “so obvious that no authority is needed” is obviously false.
MD it’s well known by US patent attorneys that a science fiction novel blocked patenting of the therapeutic waterbed. See the following article linked in the Wikipedia entry for “waterbed.”
“What science-fiction icon’s writings prevented inventor Charles Hall from patenting his designs for the modern waterbed?
Robert A. Heinlein, author of dozens of landmark science-fiction novels and stories—including Starship Troopers, The Puppet Masters, The Moon is a Harsh Mistress, Rocket Ship Galileo, and Stranger in a Strange Land. The last of these was one of the works cited by the U.S. Patent and Trademark Office as grounds to deny Charles Hall his waterbed patent in 1968.
Heinlein first described a waterbed-like invention in his 1942 novel Beyond This Horizon. He did so again in 1956 in Double Star and a third time in 1961 in the pages of Stranger.”
Very informative. Thanks.
But my reading of the Wikipedia entry for “waterbed” reveals to me that Hall despite all that made a nice pile of money out of his waterbed patents. It was just that the very broadest conceptual claims initially filed on behalf of Hall proved to be untenable against the Heinlein fantasy fiction.
Perhaps anon can now entertain us with a riff on his Ladders of Abstraction?
Do not feed Billy again until Billy has had time to digest.
Sound advice. Will keep it in mind. And thanks for the laugh.
Taking advice from Malcolm – careful of the company that you keep.
The “waterbed” example is perhaps the exception that proves the rule
Again, the emphasis is NOT that works of fiction may be absent of non-fiction items, but that ANY such non-fiction items have better sources for patent law purposes.
As to Ladders of Abstraction, please be more clear in the point that you are attempting to make. Do.
we’re all curious, Snowflake. what is the authority you cited that made the mean imaginary examiner withdraw the rejection in your imaginary response to your imaginary rejection in your imaginary prosecution at the USPTO?
Try not to say Ladders of Abstraction.
lol – your 0bsess10n is showing…
link to patentlyo.com
Ever heard the expression “Collapse of stout party” anon? As a punchline it often comes in handy and came to my mind as I read your “exception proves the rule” comment. I mean, what Rule would that be then, anon?
The expression has obscure origins in 19th century England but has endured and is handily useful even today. Perhaps then it has found its way across the pond?
Never heard of the expression.
Perhaps you have never heard of the one I used (the exception that proves the rule).
yes MD we have an expression here “i can’t come up with any authority for the dreck I made up so I’m going to go ad hominem and then try to change the subject because i think nobody will notice.”.
you just witnessed it.
and he always thinks the highly intelligent folks here never notice. There must be a word for that mental defect.
Shifty, you keep on trying to play your “gotcha” game and project things to me that just are not true.
How many years now have you been at that without even a single item of success?
Say, “ La Vee.”
Zero.
Karnak says:
“What is the value that Shifty provides?”
Are we going to have to close comments on you again, Snowflake?
^^^ retread
(You seem to have a lot of those)
Poor Snowflake. We got the last and best word in the Sonos v. Google Appeal thread and then closed comments. Snowflake has been even more discombobulated even more than usual since.
But before we close comments one again tell us again (well, first time) the authority you cited in the imaginary response to the imaginary examiner that says a sci fi reference is “improper.”
You have had the last word but a single time (the exception that proves the rule).
You have never had the best word.
Ever.
Well , that must have been persuasive to the imaginary examiner.
Did the imaginary examiner commit seppuku after such a devastating response?
[what a maroon]
Is that going to be your latest 0bsess10n?
yes call it an obsession once again asking for your authority that was so devastating the imaginary examiner committed seppuku.
Ok so cite us that imaginary authority that says a sci fi reference is “improper.”
[what a maroon]
Of course you have none. [big surprise]. We’ll just add it to your list of debacles.
Asked and answered.
(use your noggin)
Snowflake likes to parrot lawyer words when he’s really confused.
[what a maroon]
No confusion here.
Are you projecting again?
He said he made it up and had no authority for a sci fi reference being “improper” so technically Snowflake is correct asked and answered.
Is it “fiction” that you do not understand, or is it “improper” that you do not understand?
Snowflake is trying to figure the difference between some words in his head so just let him be you PEOPLE!!!
Your projection is noted.
Thank$
And now Snowflake is noting his own projections!!!
so what’s this devastating retort you gave the imaginary examiner that cited authority for the belief that sci fi literature is “improper?”
if asked and and answered as you say no need to repeat just provide the link where you answered.
[what a maroon]
Nice double down in projecting your own projections.
Plus now, you are repeating the same comments on your multiple threads of 0bsess10n with me.
On the other thread, it may have made just a little sense, as the comment that you seek was made on this one, but here it is above, and you are merely being lazy.
link to patentlyo.com
you are such a disappointment, Snowflake. we thought you had finally provided a link where you told us that devastating cite you gave the imaginary examiner in that imaginary prosecution where you established a reference to sci fi literature is , in your word, “improper.”
You are projecting again.
Thank$ for your choice.
well ok enlighten us as to this devastating retort you gave the imaginary examiner where you cited authority for the belief that sci fi literature is “improper?”
if asked and and answered as you say no need to repeat just provide the link where you answered.
[what a maroon]
What? This fractured persona of your mind was feeling left out?
There should be a difference between legitimate normal patent or publication prior art Found by AI vis a vis prior art Generated by AI that is only in some kind of alleged AI publication. The requiste authentication of the latter for litigation or IPRs may be more challengable. Especially for alleged 102 prior art references, which have an enablement requirement that 103 references do not require, as discussed before on this blog.
You make it sound like science fiction (i.e., ‘not enabled’) is perfectly valid prior art for 103 purposes.
Care to clarify?
Not to speak for Paul, but perhaps he was thinking of Symbol Techs. Inc. v. Opticon Inc., 935 F.2d 1569, 1578 (Fed. Cir. 1991): “a non-enabling reference may qualify as prior art for the purpose of determining obviousness under § 103”, which does seem rather on point.
No – that does not address the notion of science fiction as 103 material.
Thanks B, that saved me from again looking up and reminder-citing Symbol Techs. Inc. v. Opticon Inc., 935 F.2d 1569, 1578 (Fed. Cir. 1991) for “a non-enabling reference may qualify as prior art for the purpose of determining obviousness under § 103” or other such decisions.
[It would be nice if there really was a Fed. Cir. decision specifically holding that no science fiction publication suggested invention can be even part of any 103 prior art rejection even against very broad claims, since such a decision would seem to encompass what is in some AI-generated alleged prior art.]
What is the meaningful distinction between a description of a desirable useful technology in a literary work classified (by who?) as “fiction” versus a description of a desirable useful technology in a literary work classified as non-fiction. Is the description of the technology less informative because it appears in a work of fiction? Less inspiring? Less motivating? Does it matter if the work of fiction is read by tens of millions of people while the work of non-fiction was read by nobody?
The fact of the matter is that most “fiction” consists mainly of descriptions of the real world and activities in the real world that are not “fictional” at all. That’s how literature works.
In any event, it’s the actual non-fictional solution to the problem that gets patented, not the mere conception of the idea that “this thing would be cool”. Of course, recognizing this is problematic for a certain class of “inventions” that are nothing more than abstractions in the first place (“take input data and apply logic to it! am I not special?”).
I’ve covered the meaningful difference in the past.
You want elements of a science fiction reference for the non-fiction parts? By all means choose a proper non-fiction reference.
It really is that simple.
(the phrase – lacking so far that I was hoping someone would have mentioned is “for what it teaches”). Prior art must be non-fiction for it to be teaching something (anything) for the legal purposes of 103.
“ Prior art must be non-fiction for it to be teaching something (anything) for the legal purposes of 103.”
Why? Did you read this on a stone tablet you found somewhere?
Motivation to combine can arise from any source, as long as the motivation is reasonable. Logic alone should suffice. What is the argument to the contrary?
These feeble attempts to limit what is “relevant” prior art through abstract and arbitrary classifications (“this art good, that art bad!”) are never going to hold water.
I am talking about references – try to keep up.
(motivation of by way of fiction… what of ‘that it teaches’ fits?)
Me, I have difficulty distinguishing between “fiction” and a so-called “prophetic example” in a patent specification. Is there a bright line between the two forms of prior art “reference”? Anybody care to reveal it?
If you really have “AI infinite monkeys” working hard, a few of them will undoubtedly enable something accidentally. Or at least semi-plausibly enable lots of random components.
Re 103: I think the immediate problem is hindsight. AI-enabled search systems will be pretty good at stitching together the random components from the AI infinite monkeys…particularly given your applicant’s spec/claims. But, would a POSITA really have asked that question? I think the right way forward here is some combination of: (i) a limit on the number of references that can be used; and (ii) a return to TSM analysis (the Fed Circuit was right and SCOTUS was wrong).
Re 102: this is the tricker problem, though less urgent, as it’s less likely the AI infinite monkeys will randomly generate everything in your specification (OTOH, that very randomness raises fairness issues). Maybe redefine “printed publication” to require peer-review?? This would exclude AI infinite monkey garbage art, but also the old IBM technical disclosure bulletins (bad?)
I searched for prior art on ski boots on the all prior art site. This is the nonsense it came up with:
The present invention relates to a hot dip galvanized steel sheet and a manufacturing method thereof. The bag provides a transport mechanism for carrying various items, such as ski boots or snowboard boots, gloves, helmets, hats, goggles, and other snow equipment common to skiing, snowboarding or other outdoor snow sports. Another aspect of the invention provides a system and method of confirming the receipt of a transmitted data stream in a wireless local area network, wherein the system transmits a plurality of fragment packets without waiting for an acknowledgement for each fragment packet from the receiving party.
Since everything that can be invented has been invented, does any of this even matter?
Also absent from all of this are the LLM-assisted tools used to generate standard submissions and patent applications, which then automatically become prior art, but have the same basic enablement issues.
I don’t think so.
Those would fall under the category of submitted applications to which inventors and attorneys have taken oaths to.
Or are you thinking of those new apps that also submit to the patent office without any of that pesky inventor/practitioner stuff?
The USPTO continues to GRANT PATENTS on homeopathic treatments that are expressly described as such so the idea that the Office is going to be able to reliably and systematically weed out prior art that “doesn’t contribute to public knowledge” is beyond laughable.
link to patentimages.storage.googleapis.com
mm bro how you feelin today?
link to youtube.com
Nice to see 6 hooked into Krystal and Saagar.
Dennis writes: “ For example, I estimate that over the past 2 years AI systems have spit-out more text content than the total of all prior human-written publications.”
I recall you bringing this up before and I’ll ask the question again (apologies if I missed your earlier answer): Can we see the math?
I’m working on the math. Right now it is not reliable.
That’s perfect then – it matches Malcolm’s comments.
If the text is not saved to file- just generated in memory & then cleared, does that change the character of the inquiry?
I’d say being “fixed to a medium” is pertinent….
Let’s remember who among us here highlighted this problem first, mkay, and noted that it was vastly more damaging to “patent rights” than the ridiculous non-issue of “computer inventors”? That’s right: it was me. You’re welcome. It’s also nothing radically new as computers have been used to generate reams of prior art for decades now. We used to call it “poisoning the well.”
As for this:
“ For the collective, AI are creating many worthless disclosures … The difficulty lies in distinguishing between legitimate insights and spurious connections.”
Gee, why can’t we just use computers to detect the difference?
LOL
lol – your “warning” about computers has as much to do with Artificial Intelligence as you have in regard to any actual intelligence.
And that would be zero.
Thanks!
I suppose your “Why can’t we use computers…” comment was intended to be facetious but I don’t see it like that.
Here in Europe, an “invention” that is patentable has to be an enabled technical solution to a hitherto unsolved technical problem. Call your problem “X”. So, if you think you have in your possession an invention that might qualify as patentable simply go to the most powerful AI you can find, and ask it “What solutions to problem X can you find for me?”
Sit back and wait for the answer. If the output fails to include your solution, you might have not only patentable novelty but also a useful and non-obvious solution deserving of recognition as patentable.
Where’s the problem with that?
Care to provide an answer to my prior Hobson’s choice vis a vis Generative AI?
Yes. You are full of beans.
Not a cogent answer when the (Hobson) choice is of two – and only two – alternatives.
But you already knew that, eh?