By Dennis Crouch
Over the past year I’ve been investigating various generative Artificial Intelligence (GenAI) tools for assisting patent attorneys in their practice. I have a strong belief that these tools and their progeny are now fixtures in our legal environment and are being used to both improve efficient delivery of legal services and to also improve the quality of those services. Of course the generative creativity of our LLMs go hand in hand with hidden false narratives or hallucinations. Vendors are stepping up to thread the needle here: providing valuable GenAI tools while limiting false story telling. As we move forward some of the struggle will be a focus on how much the attorney needs to know about how the GenAI works in order to use it responsibly.
Enter the USPTO and its Wet Blanket: The USPTO has released new guidance on the use of AI tools in practice before the USPTO. USPTO AI Usage Guidance. The notice largely offers cautionary guidance — focusing on the need to mitigate the risks associated with the use of AI through both technical measures and human oversight. As stated in the notice, “While the USPTO is committed to maximizing AI’s benefits and seeing them distributed broadly across society, the USPTO recognizes the need, through technical mitigations and human governance, to cabin the risks arising from the use of AI in practice before the USPTO.”
One of the key takeaways from the notice is that existing USPTO rules and policies apply to the use of AI tools in matters before the Office. The notice is designed as a reminder to individuals involved in proceedings before the USPTO of that these pertinent rules and policies, and provides suggestions to mitigate the risks associated with AI use.
The duty of candor and good faith is a central focus of the guidance. The notice emphasizes that this duty applies to all conduct before the USPTO, stating that “the duty of candor and good faith applies to positions taken by applicants or parties involving the claimed subject matter.” Practitioners must ensure that any AI-assisted work product is carefully reviewed for accuracy and completeness before submission to the USPTO. It is not the AI filing work with the USPTO, but rather it is the party and their attorneys.
The notice also addresses the signature requirement and corresponding certifications under 37 CFR 11.18(b). By signing or presenting a piece of correspondence, the party certifies that all statements made therein are true to the best of their knowledge and that they have performed an inquiry reasonable under the circumstances. The notice makes the strong claim that “Simply relying on the accuracy of an AI tool is not a reasonable inquiry.”
In the context of patent applications, the notice provides specific examples of how the duty of disclosure may be implicated when using AI tools. For instance:
if an AI tool is used in drafting or editing a document, the party must still review its contents and ensure the paper is in accordance with the certifications being made. For example, given the potential for generative AI systems to omit, misstate, or even “hallucinate” or “confabulate” information, the party or parties presenting the paper must ensure that all statements in the paper are true to their own knowledge and made based on information that is believed to be true. Additionally, the party or parties should also perform an inquiry reasonable under the circumstances confirming all facts presented in the paper have or are likely to have evidentiary support and confirming the accuracy of all citations to case law and other references. This review must also ensure that all arguments and legal contentions are warranted by existing law, a non-frivolous argument for the extension of existing law, or the establishment of new law.
The guidance also addresses the use of AI tools in accessing and interacting with USPTO IT systems. Users must ensure that their use of these tools does not violate federal and state laws, as well as USPTO regulations and policies. The notice reminds users that “AI systems may not obtain a USPTO.gov account” and that “practitioners may not sponsor AI tools as a support staff individual to obtain an account.”
Confidentiality and national security considerations are also discussed in the notice. The use of AI tools may result in the inadvertent disclosure of client-sensitive or confidential information to third parties. Practitioners must be vigilant in ensuring the confidentiality of client data when using AI tools or relying on AI-related third-party services. The notice also warns that “AI tools may utilize servers located outside the United States, raising the likelihood that any data entered into such tools may be exported outside of the United States, potentially in violation of existing export administration and national security regulations or secrecy orders.” The USPTO appears to be quite serious on this front – trying to make it crystal clear that fraud and intentional misconduct will not be tolerated in any manner before the Office or in connection with accessing USPTO IT systems.
Still, the Guidance attempts to establish a middle ground, recognizing that “[p]atent practitioners are increasingly relying on AI-based tools to research prior art, automate the patent application review process, and to gain insights into examiner behavior.”
We might talk through one simple example here. With the USPTO’s upcoming fee increase for IDS filings in Fall 2024, many folks will be willing to pay to both identify the best prior art and to use automated tools to eliminate references that are either cumulative or not material to the claims. In some patent cases 10,000+ pages of prior art can easily be identified, and neither the patent attorney nor the examiner wants to individually cull through each reference. AI tools seem to offer an excellent tool for reducing the problem.
The USPTO’s guidance document raises a different concern – that the agency will be bombarded with too many references for an examiner to individually review. On this point, the guidance doubles down on prior statements that:
By signing, [the patent practitioner who submits the IDS] is certifying that they have performed a reasonable inquiry—including not just reviewing the IDS form but reviewing each piece of prior art listed on the form—and determined the paper is compliant with 37 CFR 11.18(b). Regardless of where prior art is found, submitting an IDS without reviewing the contents may be a violation of 37 CFR 11.18(b).
In my view, this approach is backwards, especially when coupled with the new IDS fees — the biggest reason is that the USPTO examiner is better off with having more prior art than less prior art. Our search tools are powerful enough for examiners to very quickly search through 10,000 pages of documents to identify those that are most relevant to the claims being pursued and then rely upon them for a rejection. This only becomes a problem when the patent office requires a human (either patent attorney or patent examiner) to actually read through each document individually.
The guidance offers other suggestions for using AI tools:
- Drafting the patent specification itself as well as responding to office actions: “For example, recent tools directed to the IP industry include the ability to draft technical specifications, generate responses to Office actions, write and respond to briefs, and even draft patent claims.”
- Conducting and reviewing prior art searches: “Patent practitioners are increasingly relying on AI-based tools to research prior art, automate the patent application review process, and to gain insights into examiner behavior.”
- Generating patent claims: “In situations where an AI tool is used to draft patent claims, the practitioner is under a duty to modify those claims as needed to present them in patentable form before submitting them to the USPTO.”
- Automating the patent application review process: “Patent practitioners are increasingly relying on AI-based tools to research prior art, automate the patent application review process, and to gain insights into examiner behavior.”
- Preparing evidence of patentability or unpatentability: “AI systems could also be used in the submission of evidence of patentability or unpatentability (e.g., evidence of secondary considerations). Though AI may be used to identify evidence or even draft affidavits, petitions, responses to Office actions, etc., practitioners are required to verify the accuracy of factual assertions, both technical and legal, and ensure that all documents, including those prepared with the assistance of AI, do not introduce inaccurate statements and evidence into the record, either inadvertently or intentionally, or omit information that is material to patentability.”
USPTO’s guidance on the use of AI tools in practice before the Office is a timely and necessary response to the growing adoption of these technologies in the legal profession. Patent law professionals, including patent attorneys and USPTO patent examiners, must familiarize themselves with this guidance and ensure compliance with the existing rules and policies when utilizing AI tools in their practice.
“This only becomes a problem when the patent office requires a human (either patent attorney or patent examiner) to actually read through each document individually.”
Back in the day, prior art searching was done by manually searching through the “shoes” of the search rooms for references. At that time, most Examiners only looked at the IDS in order to compare their search results with the references cited in the IDS. Most could not be bothered to go to the search rooms to look at all the references cited in the IDS.
“By signing, [the patent practitioner who submits the IDS] is certifying that they have performed a reasonable inquiry—including not just reviewing the IDS form but reviewing each piece of prior art listed on the form—and determined the paper is compliant with 37 CFR 11.18(b). Regardless of where prior art is found, submitting an IDS without reviewing the contents may be a violation of 37 CFR 11.18(b).”
And by signing off on an IDS, aren’t Examiners indicating that they reviewed each reference listed on the IDS? And if they didn’t actually review each listed reference, what penalties do they face? How would anyone even know if they hadn’t actually reviewed each listed reference?
“Most could not be bothered to go to the search rooms to look at all the references cited in the IDS.”
Examiners didn’t have to go to the search rooms to look at all the references cited in the IDS. The rules required that copies of the references, even of U.S. patents, had to be provided. That changed of course once all the U.S. patents and U.S. PGPUBs were available for searching through the search tools provided to the examining corps.
The point remains: The examiners did not individually look at every reference, let alone every page of every reference.
Now did they? Have they? Ever?
You have no idea what you are talking about. In the days before electronic searching, when copies of the references were required to be provided with the IDS, the examiners considered the references submitted in an IDS the same way they considered the references they searched in the shoes.
Define “consider” – verify whether or not my words are met by your words.
It’s abundantly clear that you have zero understanding of how examination is actually conducted. Your insistence on putting your ign0rance constantly on display is just beyond annoying. Even for you.
How Malcolm of you – brush away the insult and you are left with nothing.
Try to answer my question (and maybe, just maybe, you might show that the non-patent law flotsam that clogs your brain still lets in some cognitive thinking).
Still waiting for something actually meaningful and on point from you.
No you’re not. You have zero understanding of the examining process. You don’t understand how examiners search. You don’t understand how examiners consider the results of their searches, and the references that are submitted via IDS, in the process of examining the claims for novelty and non-obviousness. That is crystal clear. And you don’t want to understand. You’ve made that crystal clear with your posts here over the years. So spare me the “still waiting” nonsense because I know, and you know too, that nothing I provide in response is going to result in anything resembling an attempt to understand on your part.
Bloviate some more there, Smelly.
You think anyone is impressed that you were an examiner once upon a time — especially as you continue to dodge my simple requests?
You haven’t even tried – instead jumping to that Malcolm empty 1nsult mode (which does not work for him either).
At least try.
Didn’t mention I was an examiner to impress anybody. Least of all you.
My point stands. You have zero understanding of the examination process. And even less than zero interest in trying to understand it.
So STEFU.
Your “point” does not stand – you have provided nothing of substance.
Try again (or to be more precise, for the first time).
As typical, Smelly chooses ‘discretion’ over valor.
You’re right the references having to be provided. It’s been awhile, and I’d forgotten that references had to be provided (and at some point, they changed that to the references needed to be provided unless they’d been provided in a parent application (filewrapper continuation (FWC) application)). However, most Examiners still acted the same in that they did not review any of the references provided unless they’d independently come across them during their own search. No SPE or Primary ever inquired as to whether the Assistant Examiner had reviewed all the references listed on the IDS.
“However, most Examiners still acted the same in that they did not review any of the references provided unless they’d independently come across them during their own search. No SPE or Primary ever inquired as to whether the Assistant Examiner had reviewed all the references listed on the IDS.”
I was an examiner for 9 years and I can say this is untrue.
What time frame were you at the Office?
1990’s
Likewise. Mid-90s. I’m merely relating the way things were in my AU and among quite a few other Examiners I knew. Two things can be true at the same time.
Please Pardon Potential re(P)eat…
Your comment is awaiting moderation.
April 12, 2024 at 5:25 am
ipguy – how is it that you are simply far more reasonable than Mr. Breeze?
“However, most Examiners still acted the same in that they did not review any of the references provided unless they’d independently come across them during their own search.”
Again, not true.
“No SPE or Primary ever inquired as to whether the Assistant Examiner had reviewed all the references listed on the IDS.”
Also not true.
But now it’s, ” I’m merely relating the way things were in my AU and among quite a few other Examiners I knew.”
“Two things can be true at the same time.”
Agreed. But that doesn’t make anything you said true.
“Agreed. But that doesn’t make anything you said true.”
And we only have your word for what you say is true. Whether or not people choose to believe you (or not) is up to them, just as it is up to them to decide if they believe me to be relating the truth of my experience at the Office.
““No SPE or Primary ever inquired as to whether the Assistant Examiner had reviewed all the references listed on the IDS.”
So if even one SPE or primary inquired into whether an assistant examiner had reviewed all the references listed on the IDS, then this statement would be incorrect. Right?
Because I can remember more than one SPE and primary asking me if whether I had reviwed all the IDS references.
“Because I can remember more than one SPE and primary asking me if whether I had reviwed all the IDS references.”
And in my experience, and those of other Examiners I associated with at the time, no SPE or primary asked whether they had reviewed all the IDS references.
So if no SPE or primary inquired into whether an assistant examiner had reviewed all the references listed on the IDS, then this statement would be correct. Right?
Two things can be true at the same time.
“So if no SPE or primary inquired into whether an assistant examiner had reviewed all the references listed on the IDS, then this statement would be correct. Right?”
If no SPE or primary ever inquired, then your statement would be correct.
But SPE’s and primaries did inquire. So your statement is not correct.
Smelly, he is clearly not talking about YOUR SPE or primary.
“By signing, [the patent practitioner who submits the IDS] is certifying that they have performed a reasonable inquiry—including not just reviewing the IDS form but reviewing each piece of prior art listed on the form—and determined the paper is compliant with 37 CFR 11.18(b). Regardless of where prior art is found, submitting an IDS without reviewing the contents may be a violation of 37 CFR 11.18(b).”
But, there is no such certification on the IDS forms.
PM 2 – see below at:
link to patentlyo.com
I rather say no certification for any review is made.
Please Pardon Potential re(P)eat…
Your comment is awaiting moderation.
April 11, 2024 at 1:57 pm
Understood – but “review” is a word that simply carries no viable penalty.
As ipguy above notes, Examiners have long been hiding behind the vagaries of the word.
And the USPTO never applies a double standard… 🙂
Examiners give the references submitted in an IDS the same consideration they give references they find during their search. It’s not a difficult concept to understand. Unless you’re being willfully obtuse.
“Consideration”
And you want to label others obtuse?
Really?
I don’t think ipguy was ever an examiner and whoever is feeding him information apparently doesn’t know anything about how examiners treat IDS’s either.
“treat”
Wow – what a treat.
When an examiners signs off on an IDS that is the examiner’s certification that the references have been considered.
“Considered”
Sure.
But what does that entail? Precisely?
Oh wait – this has already been put to you for another term.
Mr. Crouch,
Your statement “Our search tools are powerful enough for examiners to very quickly search through 10,000 pages of documents to identify those that are most relevant to the claims being pursued and then rely upon them for a rejection. ” is not wholly correct. For patent examiner’s if the document is a US pat or PgPub then our tools will allow for one to search these with key words. However, if they are a foreign patent, or god forbid, an NPL then there is no easy way to search. This is because they are presented to as a PDF. And, there is no quick way to search. Therefore, your statement is partially incorrect. More is not always better, in fact, generally speaking more is often worse as there is no way for an examiner to wade through the thicket of references to determine what is actually applicable to the claims. In addition, examiner’s have to always consider all references. Which generally means examiners have to read the reference. Having 10,000 pages of references is a huge burden upon examiners as the applicant do not have to provide any sort of guidance on what part of the 10,000 pages is actually relevant to the examination of the claims.
The office has extremely powerful scan and search tools that they are not providing to examiners.
Indeed. And that may be prudent considering how the office is currently failing to provide basic search tools to examiners on a consistent basis. Let’s not distract management from the need to provide examiners with the ability to keyword and classification search during work hours.
Does this have to do with conducting examination without public disclosure?
Talk about putting yourself out of a job!
… and yes, these points reflect my own past calls for discussions of things AI, that I first presented years ago in relation to the DABUS case.
Of course, my detractors will again deny this.
I’ll bite. Who cares enough about you to be a detractor?
Does this mean you don’t want to buy an “I’m an anon detractor” T-shirt? One size fits all.
^^^ is that like the magical one-size-fits-all Malcolm bucket?
^^* you bite…
… yourself.
Thank you for proving my point.
(Sheesh, litigators)
You had a point? Are you saying that your meandering rants were intended to have a purpose other than humor? Whoa!
NS II – way to double down.
You really are only a litigator, aren’t you?
Since everything is irony, you are doubtless your biggest detractor.
That’s likely the reason for your baroque overcompensation.
But also since most people are mostly decent, the likely predominant response is a kind of sympathy.
As Emperor Tang says: “Sad”
… and its Opposite Day here at the ‘O.
Connect the dots:
Combine this with the prior notice requesting differentiation and identification of any inventive aspects that distinguishes what a human – in and of themselves – may rise to be a legal inventor, and this is a clear warning that any inventions that do not rise to being fully and totally (only) by humans, will not be enforceable.
… this will be a wet blanket for a certain Pharma executive who has testified publicly that their inventions were made by AI (just about a year ago, as I recall).
“…(just about a year ago, as I recall).”
Cite? Maybe you can find it along with that “credible source” that reported that 95% of all lawyers are “hard left.”
Lulz
Your lulz is rather odd.
Especially given your own lack of substance on most all things non-patent.
Lovely glass shards of your former house that you are standing in.
Will examiners have a duty to disclose their use of AI?
NOIP,
I would be interested in verification from the Patent Office that their admitted use (already) of AI tools is done in a manner required under law of not divulging anything to the public.
“In my view, this approach is backwards, especially when coupled with the new IDS fees — the biggest reason is that the USPTO examiner is better off with having more prior art than less prior art. ”
Rule 56 should just be done away with.
As for IDS, the key word of “review” (as in, review every document) is rather conveniently undefined and something as simple as “Yup, that’s a document, go ahead and include it” suffices.
For what it’s worth, most (all?) of these “guidelines” applied to (1) using a computer to search the Internet and copying/pasting information found there into a patent application, and (2) using spell/grammar checking programs and/or application templates discovered online.
Computers are t 0 0 ls. Use wisely.
lol – how did your attempt at reformatting your hard drive with that Britney Spear CD go for you?
Your attempted equivalence statement here is on par.
Associates will be motivated to increasingly rely on AI tools to maximize their time/billing stats while providing little/no oversight over the AI tool output.
The way most partners review associate work product makes it unlikely that hallucinations/inaccuracies will be caught.
Patent Examiners won’t catch these either, leaving these ticking timebombs for the client to discover in litigation where opposing counsel will most likely find the AI tool hallucinations/inaccuracies.
Of course this will all be blamed on “the system,” leaving clients holding the bag.
+1 I can imagine a death spiral for the profession: current cost pressure –> use of AI –> “patents aren’t enforceable anymore” –> more cost pressure.
New associates should probably focus on the litigation track going forward. Or maybe something safer, like trademarks.
Computers are t 0 0 ls.
That is correct. They are not recipes.
And changing the programming of that computer can completely change the functionality of that tool. And tools are generally considered patentable subject matter.
Thanks for the admission!!!
B-b-b-but that’s only abstract plus old computer…
(Invitation to the Grand Hall experiment yet to be redeemed)
… also (per Malcolm), three resistors are the same as three resistors.
Doesn’t matter at all that one set of three is in parallel and another set is in series.
“tools are generally considered patentable subject matter”
Not when they are old tools. When was the programmable digital computer invented?
2013, right?
lol – the vapid television show “House” answer — and immunized against with the Grand Hall experiment and the three resistors retort.
You stepped right into it, Malcolm.
Nice job
… you also open the door to the “old items of protons, neutrons, and electrons.”
Further, your “Ends justifies the Means” trope is exposed as “old” has nothing to do with “eligible” as separate patent law concepts.