NSCAI Final Report: United States Must Up Its IP Game to Win the AI Race

By Jennifer Maisel
March 11, 2021

“Significant questions remain as to whether U.S. courts will reliably permit IP holders to proceed with AI-focused IP infringement claims against potential offenders, or if patent-eligibility, inventorship, fair use, and other defenses will tip the scales towards trade secret protection.”

https://depositphotos.com/162980480/stock-photo-cyborg-running-fast.htmlThe National Security Commission on Artificial Intelligence (NSCAI) recently issued its Final Report outlining a strategy for the United States to “win” the artificial intelligence (AI) era. Critically, the Commission argues that the United States government is not currently organizing or investing to win the technology competition in AI against a committed competitor (i.e., China), nor is it prepared to defend against AI-enabled threats and rapidly adopt AI applications for national security purposes. In addition to the potential patent-eligibility and data ownership IP issues noted in the Report, other IP uncertainties unique to AI technology continue to persist, such as machine authorship and machine inventorship of valuable contributions. As part of the United States’ overall strategic interests in winning the AI era, the Commission recommends that the United States adopt IP policies to incentivize, expand, and protect AI and emerging technologies, as well as recognize IP as a national priority. But significant questions remain as to whether U.S. courts will reliably permit IP holders to proceed with AI-focused IP infringement claims against potential offenders, or if patent-eligibility, inventorship, fair use, and other defenses will tip the scales towards trade secret protection.

[[Advertisement]]

Uncertainties in IP Policy for AI Innovation

In chapter 12 of the Report, the Commission highlights how stringent patent eligibility requirements in U.S. courts and a lack of explicit legal protections for data have created uncertainty in IP protection for AI innovations, particularly with respect to patents.

By way of background, the United States Patent and Trademark Office (USPTO) revamped the patent eligibility requirements in 2019, resulting in a 44% decrease in uncertainty of patent examination concerning subject matter eligibility in the one year following the 2019 Revised Patent-Eligibility Guidance. U.S. courts, however, have only recently begun applying patent eligibility requirements to AI inventions—thus far with mixed results and continued uncertainty. See, e.g., Quad City Patent, LLC v. Zoosk, Inc., No. 20-cv-01996-JST (N.D. Cal. Nov. 3, 2020) (Opinion) (finding claims ineligible under Section 101). Accordingly, further guidance from the Supreme Court or a legislative effort to reform Section 101 may be necessary to ensure that AI innovators can not only obtain patent protection, but reliably enforce their patent rights in a U.S. court.

Moreover, the Commission’s observation regarding a lack of explicit legal protections for data echo recent sentiments from industry, as reflected in the USPTO’s October 2020 report on Public Views on Artificial Intelligence and Intellectual Property Policy. The USPTO observed that commentators “were nearly equally divided between the view that new intellectual property rights were necessary to address AI inventions and the belief that the current U.S. IP framework was adequate to address AI inventions.”  While an appeal is currently pending in the U.S. District Court for the Eastern District of Virginia, last year the USPTO held that an inventor to a patent must be a natural person. The Copyright Office has similarly required human authorship for a copyrighted work. See, e.g., Compendium of U.S. Copyright Practices, § 306 (“The Human Authorship Requirement”). In addition, significant questions about the scope of copyright protection for certain types of software (i.e. “Application Programming Interfaces”) and the bounds of a fair-use defense are currently pending before the Supreme Court (Google LLC v. Oracle America Inc., No. 18-956). Accordingly, current IP policy potentially leaves certain discoveries and data output from machines outside the realm of patent or copyright protection.

Findings and Recommendations in the NSCAI Report

In the Report, the Commission notes that the “U.S. has not developed comprehensive IP policies to incentivize investments in and protect the creation of artificial intelligence (AI) and other emerging technologies,” and that China is “poised to ‘fill the void’ left by weakened U.S. IP protections, particularly for patents.”  The Report highlights the following key areas that have weakened U.S. IP protections:

  • S. Courts have severely restricted what types of computer-implemented and biotech-related inventions can be protected under U.S. patent law;
  • China has met its strategic policy goal of increasing the quantity of its patent applications and issued patents, creating the narrative that it has “won” the innovation race;
  • China’s prolific patent application filings have created a vast reservoir of prior art that now must be reviewed in examining a patent application;
  • China’s companies have been identifying many patents as “standard-essential” in standards development organizations;
  • The lack of explicit legal protections for data or express policies on data ownership may hinder innovation and collaboration, particularly as technologies evolve; and
  • China continues to pervasively steal American IP-protected technological advances through varied means like cyber hacking of business and research institutes, technological espionage, blackmail, and illicit technology transfer.

To fill the IP policy void, the Commission recommends that the “United States must, at a minimum, articulate and develop national IP reforms and policies with the goal of incentivizing, expanding, and protecting AI and emerging technologies, at home and abroad.”  For next steps, the Commission recommends that the President issue an executive order recognizing IP as a national priority and directing the Vice President, as chair of the Technology Competitiveness Council, or an interagency task force to oversee this effort. The Secretary of Commerce, in coordination with the Under Secretary of Commerce for Intellectual Property and the Director of the USPTO, should further be directed to develop proposals to reform and establish new IP policies and regimes, including proposals for executive and legislative actions and an assessment of a non-exhaustive list of IP considerations.

A Patent Upswing for AI

AI innovators should take note that the Report–consistent with other executive federal policy concerning AI—appears to favor forms of IP protection that promote the disclosure of AI technology to the public, such as through a patent filing, rather than keeping the AI technology a trade secret. Indeed, as reflected in the USPTO’s October 2020 report on Inventing AI, the number of AI-focused patent filings has exploded over the past several years: “One hallmark of valuable new technologies is an increase in patent applications. These applications reflect the expectations and decisions of investors and innovators who seek to use or build on the new technologies for innovation. AI technologies exhibit this increase.”

Image Source: Deposit Photos
Image ID:162980480
Copyright:phonlamai 

The Author

Jennifer Maisel

Jennifer Maisel , a partner at Rothwell Figg in Washington, DC, is a technology lawyer and an emerging thought leader on the intersection of AI and the law. Her practice encompasses all aspects of IP and privacy law issues involving cutting-edge technology. She serves a diverse range of clients – from solo inventors and start-ups to Fortune 100 companies – in matters concerning AI and machine learning, telecoms systems, the Internet of Things, Big Data technologies, blockchain, mobile and website applications and other digital technology. She has been selected to the Washington, DC Super Lawyers ‘Rising Star’ list (2018-2020) and is included in the inaugural edition of Best Lawyers: Ones to Watch (2021), which recognizes extraordinary lawyers who have been in private practice for less than 10 years.

For more information or to contact Jennifer, please visit her Firm Profile Page.

Warning & Disclaimer: The pages, articles and comments on IPWatchdog.com 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 IPWatchdog.com. Read more.

Discuss this

There are currently 14 Comments comments.

  1. Pro Say March 11, 2021 9:13 pm

    Without the restoration of patent eligibility to all areas of innovation — as this excellent study also makes crystal clear must happen — the future of American innovation looks bleak indeed.

    With the CAFC and SCOTUS continuing to unconstitutionally crush our Country’s innovation — including the innovations critical to America including the personal and financial security of its 330 million people — it is left to Congress to make right what which these courts have made so very wrong.

  2. Pro Say March 12, 2021 10:27 am

    Courtesy of this morning’s L.A. Times:

    China Delegates OK Plan to Surpass U.S.

    Beijing — In smugly celebratory tones, China wrapped up political meetings . . . and surpass the U.S. in technology.

    https://enewspaper.latimes.com/desktop/latimes/default.aspx?edid=a0c543e1-f1d1-4454-98d6-cfae189d3532

  3. Francis G Rushford March 18, 2021 11:10 am

    The US lost the AI race and the technology race years ago. It started before Patent Reform, but Patent Reform put a fork in it, along with the view that online advertising and games on phones were break-through innovations.

  4. David Stein March 22, 2021 10:27 pm

    > inventorship

    Not this again… didn’t the USPTO’s public comment period make it clear that this is currently a non-issue? From the USPTO report:

    “The majority of commenters responding to this question reflected the view that there is no need for revising patent laws and regulations on inventorship to account for inventions in which an entity or entities other than a natural person contributed to the conception of an invention.”

    To repeat my answer from the last several go-arounds with this issue:

    Inventorship involves two distinct purposes:

    Purpose #1: Documenting the creation process – i.e., identifying the people who conceived the invention and reduced it to practice.

    Expanding this purpose to allow ML algorithms to be “named” as inventors is harmless, uncontroversial, and trivial. Currently, any such patent could (and likely should) document the contribution of the algorithm to the production of inventive subject matter, simply in the body of the disclosure (or “specification,” as PatO readers appear to prefer). Whether the human inventors consider the ML algorithm to be a co-inventor or a tool is really a subjective decision, best left up to them.

    Purpose #2: Assigning ownership shares of the invention to the inventors.

    This is extremely problematic for a multitude of reasons, most of which boil down to: “algorithms cannot legally own anything, including an ownership stake in intellectual property.”

    Is this legally discriminatory? Let’s defer that discussion until an algorithm demonstrates enough AGI to raise that complaint.

  5. Anon March 23, 2021 5:34 pm

    On the contrary, Mr. Stein, I would prefer to address the problem BEFORE there is an immediate need, or perhaps more accurately put, before the appreciation of the immediacy of a need – as the need is most likely to exist in reality before we recognize it, and waiting too long is only guaranteed to generate sub-optimal reasoning and results.

  6. Anon March 23, 2021 5:38 pm

    I would also posit that AI does NOT affect ‘inventorship’ concerns alone, as it is quite easy to see how Persons Having Ordinary Skill In The Art (inclusion of AI or not) presents an incredibly difficult problem.

    Quite in fact, there is some serious legal misperceptions already existing due to the type of “the problem is not here yet” thinking.

    This aspect of inventorship is NOT about the Singularity or any complete artificial intelligence – as is often seen in cinema. The level (limited as it is to be properly appreciated) that invokes “Devisor” level status simply does not require the “whole kit and kaboddle.”

  7. David Stein March 24, 2021 11:05 am

    Anon, AGI has the same problem as the personal jetpack: it’s been “10 years away” for 80 years.

    Also, the gains in AI/ML techniques over the last decade are orthogonal to AGI. We are getting really good at using AI/ML to solve a broader range of problems, including mathematically complex problems (e.g., complex systems of Bernoulli flow equations) and poorly-defined problems for which we can’t provide full instructions (e.g., broad-spectrum anomaly detection). None of those advances truly move the ball toward AGI any more than the expanded computational features of quantum cryptography make quantum computing “generally intelligent.”

    So addressing the legal issues of AGI ownership of intellectual property today is kind of like passing laws about commercial zoning of property on Mars. It’s far too early to approach the issue in any way that is likely to make sense once the issue finally becomes relevant.

    > Persons Having Ordinary Skill in the Art

    Okay, but we can deal with the issues on a case-by-case basis. In fact, we already do – note the AIA’s expanded definition of “Applicant” to include corporations that drive, and own, the lion’s share of patents. But, again, the resolution of such existing, immediate problems should not be expanded to resolve far-future problems where we don’t even know what those far-future problems would look like.

  8. Anon March 24, 2021 3:22 pm

    Sorry but I have to disagree David – you appear to be looking at a level of “the Singularity,” and as I have noted, such is NOT the level that will be the issue (revisit the DABUS UK case in which there it was found that DABUS was the Devisor.

    This has zero in common with any notion of personalized jet packs.

  9. Anon March 24, 2021 5:05 pm

    … an example of a “right now” issue may be seen as related to the DABUS case.

    In that case, NO prosecution (anywhere) advanced for failure at the start.

    Whether or not you agree — let’s take for the sake of argument that the UK version (Devisor established) as a starting point.

    Now imagine, that instead of being the Entire inventive entity (as was being pursued), that instead a co-inventor status was the staking horse.

    What then?

    On top of that, if I recall correctly, one of the organizations recently put out a paper differentiating between “just use” (akin to NON-AI computing), and the more DABUS-like non-human appreciation of a distinction in the first instance.

    How many current items are of the nature that AI involvement (and here, for argument’s sake, incorrectly) classified as the “just use” Type I when in reality, there is SOME sliver of “Devisor” — in a joint inventor sense — that may (far later, as in litigation and enforcement time) be brought to bear?

    Ignoring this possibility now runs the very real (and not negligible) risk that “co-inventor status” must be attributed to the AI as Devisor as opposed to ‘mere use, in the way of non-AI.”

    As to the “creep” of co-inventor,” see

    https://patentlyo.com/patent/2021/03/level-contribution-inventorship.html

  10. David Stein March 25, 2021 2:54 am

    I enjoy this discussion, but I have to take issue with this point:

    > differentiating between “just use” (akin to NON-AI computing), and the more DABUS-like non-human appreciation of a distinction

    You are assuming that the algorithm has the ability to “appreciate” things, as in: not just calculating results but developing opinions about them, based on a set of personally developed criteria concerning quality and value.

    This would be an extraordinarily advanced set of capabilities, combining several traits of AGI. Nothing leads me to believe that the ML algorithms of today, or any time in the near future, will have such capabilities.

    Of course, the literature about DABUS implies that it has such powers. To date, both the actual mechanics of DABUS and any objectively reviewable results are withheld from view. The applications (U.S. App. 16/524,350 and16/524,532) both remain unpublished – which is interesting, given that both appear to have been filed in mid-2019 and can only remain unpublished due to the filing of a nonpublication request… which would appear to contradict reports of parallel applications in the EPO and UKIPO…?

    Anyway, my point is that our sole source of information about the sophistication of DABUS is the inventor’s self-laudatory publicity about it. I remain skeptical.

    Look – the history of AI/ML is replete with examples of people looking at the output of algorithms and attributing significantly more “intelligence” to them than is warranted by extended experience. The contemporary example is GPT-*. GPT-1 provides plausible output for a few sentences, at which point it devolves into obvious gibberish. GPT-2 pushes the theshold to two or three paragraphs before exhibiting the same failure. GPT-3 probably moves the threshold back further, but the threshold remains.

    Neural style transfer is the same. Lofty statements are made about “GANs are capable of *learning* the style of a famous artist!” No, they merely learn to graft simple visual patterns or sequences onto new works. Continued experimentation with GANs demonstrates that the underlying mechanism is quite primitive.

    Notably, in the case of both GPT and neural style transfer, advocates of the algorithms often support their optimism by cherry-pick great examples from a pile of duds. There is no indication that this is not the case here. The literature about DABUS gets maximally vague and “hand-wavey” around the “critic” part of the model, and instead devolves into a testy and unsubstantiated denial that DABUS is not merely a GAN.

    Time will tell, anon. The history of AGI – the cyclic issuance of overly generous assessments of the state of the art, followed by dialing-back and disappointment – is on my side. And such metaphysical conjecture has the unfortunate side-effect of obscuring *actual* developments in AI/ML, which are rapid, impressive, and widely applicable.

  11. Anon March 25, 2021 6:44 am

    Your issue with the differentiation is noted.

    In reply, I will note that it was a technical society that published that differentiation, so your beef is more than just with me.

    As to DABUS and what is and is not revealed, it is odd that “non-publication” would remain in effect because there most definitely are foreign equivalents – and those foreign equivalents have been discussed dating back more than a year now.

    Perhaps a simpler explanation might be that the applications simply have not been processed because the attempt to claim (as single and only inventor) a non-human.

    Be that as it may, your unwillingness to not engage on the issue has no actual bearing on the importance of the issue, and merely deprives us of your participation.

    I think that your characterization is simply off, and that you remain mired in thinking that a larger degree of self-sufficiency (The Singularity) is what is needed to precipitate discussion, and such is just not so. All that it takes is that the legal definition of innovator (or to springboard from the UK case for the obvious reason of an extant holding there — of Devisor) need be considered.

    I do find it odd, seeing as you seem to understand that innovation itself is forward looking, and that our craft includes drafting legal documents that seek a forward sense of protection some twenty years out, your steadfastness in clenching tight your eyes on this particular issue.

    After all, non-AI analogs DO exist (patents and slaves, simian selfies, and the like).

  12. David Stein March 25, 2021 7:41 pm

    > Perhaps a simpler explanation might be that the applications simply have not been processed because the attempt to claim (as single and only inventor) a non-human.

    Possible, I suppose. It’s my sense that the Office of Patent Application Processing (which handles ministerial functions, including pre-grant publications) does not question the substance of submissions. For example, if you submit an application with an inventor list of Elvis Presley, John Lennon, and Freddy Mercury, they would process and publish it in that form (although you would likely be referred to a federal prosecutor for a fraudulent submission). So I don’t know that they would hold an application from publication over this kind of question.

    > your unwillingness to not engage on the issue

    But I believe that I *am* “engaging on the issue.” In fact, I’ve been more engaged than most of the patent community – which has simply said “NO” – because I believe that splitting “inventorship” into a “manner of creation” component and an “ownership” component, AI/ML algorithms could be easily included in the first part.

    But if you mean “helping to find a way that AI/ML algorithms can be named as owners,” then you’re right, I will not be helpful in that way. And not because I “refuse to engage,” but, rather, because I believe that such a proposition is fraught with legal, moral, and practical peril, in part due to the very early stage of development of AI/ML. That is, I “engage” by identifying the dangers with this course of action – which, of course, is one of the principal duties of attorneys!

    > All that it takes is that the legal definition of innovator (or to springboard from the UK case for the obvious reason of an extant holding there — of Devisor) need be considered.

    Okay, but the onus is on the people who believe that the standard should change to come up with a specific suggestion. The one suggestion on the table (“ownership by algorithms”) is problematic, for reasons I can explain in more detail if you like. If there is a different suggestion, I’ll be happy to consider that one as well. But I wouldn’t be able to propose any novel suggestions, simply because I see no problem with the current standard.

    In short, I’m not refusing to engage and I’m not just being a naysayer – I’m calling hits and misses like I see them. Unfortunately, this one misses by a wide margin.

    > non-AI analogs DO exist (patents and slaves, simian selfies, and the like).

    Slaves were 100% people. The legal issue there was overturning an offensive and wrong-headed legal principle that they were less than whole people.

    Based on the Wikipedia entry, the “simian selfies” case was settled without a definitive and comprehensive answer from SCOTUS or any other court – but both District Judge Orrick and the Ninth Circuit made preliminary rulings that “animals have no legal authority to hold copyright claims.” (Obviously, “animals” meaning “non-human animals,” in these instances.)

  13. Anon March 26, 2021 5:34 pm

    But I believe that I *am* “engaging on the issue.”

    I know that THAT is what you are saying.

    But you are not doing so.

  14. George April 10, 2021 3:17 pm

    @ Francis G Rushford #3

    Agree!