[Senate Hearing 118-65]
[From the U.S. Government Publishing Office]
S. Hrg. 118-65
ARTIFICIAL INTELLIGENCE AND
INTELLECTUAL PROPERTY_PART I:
PATENTS, INNOVATION, AND COMPETITION
=======================================================================
HEARING
before the
SUBCOMMITTEE ON INTELLECTUAL PROPERTY
of the
COMMITTEE ON THE JUDICIARY
UNITED STATES SENATE
ONE HUNDRED EIGHTEENTH CONGRESS
FIRST SESSION
__________
JUNE 7, 2023
__________
Serial No. J-118-19
__________
Printed for the use of the Committee on the Judiciary
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
U.S. GOVERNMENT PUBLISHING OFFICE
53-115 PDF WASHINGTON : 2024
COMMITTEE ON THE JUDICIARY
RICHARD J. DURBIN, Illinois, Chair
DIANNE FEINSTEIN, California LINDSEY O. GRAHAM, South Carolina,
SHELDON WHITEHOUSE, Rhode Island Ranking Member
AMY KLOBUCHAR, Minnesota CHARLES E. GRASSLEY, Iowa
CHRISTOPHER A. COONS, Delaware JOHN CORNYN, Texas
RICHARD BLUMENTHAL, Connecticut MICHAEL S. LEE, Utah
MAZIE K. HIRONO, Hawaii TED CRUZ, Texas
CORY A. BOOKER, New Jersey JOSH HAWLEY, Missouri
ALEX PADILLA, California TOM COTTON, Arkansas
JON OSSOFF, Georgia JOHN KENNEDY, Louisiana
PETER WELCH, Vermont THOM TILLIS, North Carolina
MARSHA BLACKBURN, Tennessee
Joseph Zogby, Chief Counsel and Staff Director
Katherine Nikas, Republican Chief Counsel and Staff Director
Subcommittee on Intellectual Property
CHRISTOPHER A. COONS, Delaware, Chair
MAZIE K. HIRONO, Hawaii THOM TILLIS, North Carolina,
ALEX PADILLA, California Ranking Member
JON OSSOFF, Georgia JOHN CORNYN, Texas
PETER WELCH, Vermont TOM COTTON, Arkansas
MARSHA BLACKBURN, Tennessee
James Barton, Democratic Chief Counsel
Seth Williford, Republican General Counsel
C O N T E N T S
----------
JUNE 7, 2023, 3:01 P.M.
STATEMENTS OF COMMITTEE MEMBERS
Page
Coons, Hon. Christopher A., a U.S. Senator from the State of
Delaware....................................................... 1
Tillis, Hon. Thom, a U.S. Senator from the State of North
Carolina....................................................... 3
prepared statement........................................... 34
WITNESSES
Witness List..................................................... 33
Abbott, Ryan, professor of law and health sciences, University of
Surrey School of Law, Guildford, United Kingdom, and adjunct
assistant professor, David Geffen School of Medicine,
University of California, Los Angeles, California.............. 8
prepared statement........................................... 42
Elluru, Rama, senior director, society and intellectual property,
Special Competitive Studies Project, Washington, DC............ 11
prepared statement........................................... 53
Salsberg, Corey, vice president and global head of IP affairs,
Novartis, Washington, DC....................................... 4
prepared statement........................................... 60
Sheridan, Laura, head of patent policy, Google, New York, New
York........................................................... 6
prepared statement........................................... 75
Villasenor, John, professor of law, electrical engineering, and
public policy, and founder and faculty co-director, Institute
for Technology, Law, and Policy, University of California, Los
Angeles, California............................................ 9
prepared statement........................................... 81
QUESTIONS
Questions submitted to Ryan Abbott by Senator Tillis............. 89
Questions submitted to Rama Elluru by Senator Tillis............. 93
Questions submitted to Corey Salsberg by Senator Tillis.......... 97
Questions submitted to Laura Sheridan by Senator Tillis.......... 101
Questions submitted to John Villasenor by Senator Tillis......... 105
ANSWERS
Responses of Ryan Abbott to questions submitted by Senator Tillis 109
Responses of Rama Elluru to questions submitted by Senator Tillis 114
Responses of Corey Salsberg to questions submitted by Senator
Tillis......................................................... 121
Responses of Laura Sheridan to questions submitted by Senator
Tillis......................................................... 133
Responses of John Villasenor to questions submitted by Senator
Tillis......................................................... 138
ARTIFICIAL INTELLIGENCE AND
INTELLECTUAL PROPERTY--PART I:
PATENTS, INNOVATION, AND COMPETITION
----------
WEDNESDAY, JUNE 7, 2023
United States Senate,
Subcommittee on Intellectual Property,
Committee on the Judiciary,
Washington, DC.
The Subcommittee met, pursuant to notice, at 3:01 p.m., in
Room 226, Dirksen Senate Office Building, Hon. Christopher A.
Coons, Chair of the Subcommittee, presiding.
Present: Senators Coons [presiding], Hirono, Padilla,
Tillis, and Blackburn.
Also present: Senator Blumenthal.
OPENING STATEMENT OF HON. CHRISTOPHER A. COONS,
A U.S. SENATOR FROM THE STATE OF DELAWARE
Chair Coons. I'd like to call this hearing to order. Thank
you to all of our witnesses for participating today. I'd like
to thank Ranking Member Tillis and your staff for working so
well and closely with mine to put together this hearing on a
consensus basis. As you can tell from the attendance, this is a
topic of wide interest. And so to my colleague and friend,
Senator Hirono, I am looking forward to a robust hearing on the
topic of artificial intelligence and intellectual property.
We're going to explore some of the recent developments in
AI, and in particular, patents and the impact of AI and
innovation and U.S. competitiveness. Make no mistake, AI
presents novel questions across a wide range of areas of IP
policy. And today's the first in a series of hearings that this
Subcommittee will have looking at the intersection of
artificial intelligence and intellectual property law and
policy.
We've all heard about different tools like ChatGPT or
DALLE, impressive new generative AI tools that have opened up
so many creative avenues and raised many concerns. These tools
are just a few of many popular uses, but generative AI has
already been put to work to help solve a wide range of very
serious and substantive problems. Drug development is just one
particularly compelling example.
Over the last few decades, antibiotic discovery and
developments have become increasingly difficult and expensive.
Many researchers have given up because new antibiotic drugs
have not been able to overcome the threat of antibiotic
disease. In fact, no antibiotic classes have been introduced
since the 1980s.
Enter AI. Researchers at MIT recently trained an AI model
using a large collection of diverse molecules and then used the
model to make a new and potent antibiotic that is effective
against antibiotic-resistant bacteria. This discovery didn't
take years. It didn't take weeks. It took a few days.
I share this story because many of the conversations around
AI focus on the potentially harmful uses of the technology,
threats of misinformation, and bias. Highlighting these uses is
important, but I believe we also have to act thoughtfully to
set critical safeguards. It's important to shine a light on how
AI is being used to innovate. Drug development is one
compelling example. Others include its role in efforts to
combat climate change, to address our computer chip shortage,
and to create renewable energy sources.
These innovations raise new, interesting, and complex
patent law issues, including whether innovations facilitated by
AI are, or should be, patentable, and if so, who should be
listed as the inventor. Currently, in the U.S., many AI-
generated inventions are not patentable because the Supreme
Court has determined the law does not permit including a non-
human inventor on a patent application.
The decisions we make in Congress about whether and how to
protect AI-related innovations will also have significant
consequences for U.S. innovation and competitiveness. We need
to ensure we establish a rights regime that encourages AI-
generated innovation to stay here in the United States, instead
of incentivizing innovators to turn to other countries with
more favorable laws to protect their AI-generated inventions
and other emerging technologies.
Despite the critical role of IP in AI innovation, IP
considerations have largely been missing from proposed AI
regulation frameworks. Neglecting IP in an effort to regulate
AI would have serious unintended consequences for our
innovation ecosystem, our national security, our economic
competitiveness. In contrast, competitors, like China,
recognize IP policies as an important tool in national
strategies for AI and other emerging technologies.
China has elevated the role of its patent and trademark
office and has even been exploring data rights as a new form of
IP protection for AI. So what can we do? There are some initial
steps we can take to ensure U.S. leadership on AI policies. I
think it's critical that we include IP considerations in
ongoing AI regulatory frameworks and make certain the U.S.
Patent and Trademark Office has a seat at the table.
We should change our patent eligibility laws, and I suspect
my colleague agrees, so that we can protect critical AI
innovations. And last, we should consider whether other changes
to our patent laws or new and unique protections may be
necessary to encourage innovation in AI and emerging
technologies. I am excited to explore these issues with you
today. With Senator Tillis' cooperation, we have a superb panel
with a diversity of views and perspectives. You will, I
suspect, find 5 minutes a bracingly short period in which to
explain them.
We are grateful for your lengthy, submitted written
testimony and look forward to your individual summaries. But
first, I'll turn it over to Senator Tillis before I introduce
our panel. Senator.
OPENING STATEMENT OF HON. THOM TILLIS,
A U.S. SENATOR FROM THE STATE OF NORTH CAROLINA
Senator Tillis. Thank you, Chairman Coons. And thank you
for really another great hearing topic and opportunity. Before
I make a few brief remarks, I'm just curious, could anybody--if
you're under the age of 35 out there, would you raise your hand
right now?
[Audience members raise their hands.]
Senator Tillis. Okay.
[Laughter.]
Senator Tillis. So I kind of bring up the average a little
bit, but even this old guy is pretty interested in this
subject. But I thank you all for being here. I think this is
one of an endless number of hearings that we're going to have
to have to make sure that we get it right.
And it's great to highlight the importance of intellectual
property. Within about 24 hours of ChatGPT offering its beta
site, I signed up, I had--back in the mid 1980s, I was actually
working in AI when the state-of-the-art was character
recognition, pattern recognition, and voice recognition. So
we've come a long way since then, and as the tools continue to
explode, the challenges are going to be great. So this is a
good starting point.
I think the one topic--everybody's talking about all the
bad things that could happen with AI, ``I, Robot,'' the zombie
apocalypse, and the good things that are going to happen
category. And even if we chose not to participate as a nation,
it's going to happen. So the question is whether or not the
United States is going to maintain a leadership position in
innovation and what I think is one of the most promising spaces
for future innovation that we have today.
But we're having this hearing. AI is going to have a
massive impact on intellectual property. And I think this
hearing begins to allow us to start parsing this up into issues
related to patents, into issues related to copyrights, and any
number of other subjects. So I'm going to submit my opening
statement for the record. But the key here is, if anyone on the
panel thinks that AI is bad and we got to avoid it, then we'll
agree to disagree. But I'd be interested in your comments.
What I'm most interested in is seeing how this new tool,
this new capability, is going to continue to make the United
States of America the leading--the historic, leading innovator.
And I look forward to your comments.
[The prepared statement of Senator Tillis appears as a
submission for the record.]
Chair Coons. Thank you, Senator Tillis. I'm now going to
turn to an introduction of our witness panel. We welcome five
witnesses to testify about the intersection of AI and patents,
innovation, and competition.
Our first witness is Corey Salsberg, a vice president,
global head of IP affairs at Novartis, a global healthcare
company. He's responsible for defining and developing Novartis'
core positions and directing its advocacy efforts with respect
to IP law and policy.
Next, we have Laura Sheridan. Ms. Sheridan is head of
patent policy at Google and advocates for an effective patent
examination process, a patent litigation system that operates
fairly for all, and transparency in these areas. She's part of
the IP Owners Association, a delegation to IP5 Industry, and a
group which engage--a group which engages the IP5 Offices on
procedural harmonization matters.
Next is Ryan Abbott, professor of law and health sciences
at the University of Surrey and an adjunct professor of
medicine at UCLA. Professor Abbott is the author of ``The
Reasonable Robot: Artificial Intelligence and the Law.'' He
also leads the Artificial Inventor Project, representing Dr.
Stephen Thaler in the global legal test cases seeking IP rights
for AI-generated inventions.
Next, we'll hear from John--Professor John Villasenor.
Professor Villasenor teaches law, electrical engineering, and
public policy at UCLA, where he founded the UCLA Institute for
Technology, Law, and Policy. His work considers the
intersection of technology and law in areas including
artificial intelligence, digital privacy, and the technology
supply chain. He's written on the implications of AI in
relation to patents.
Finally, we have Rama Elluru, senior director for society
and intellectual property at the Special Competitive Studies
Project. She leads a team working on shaping AI governance and
IP policy. Previously served as staff to the National Security
Commission on Artificial Intelligence, she led the NSCAI's IP
efforts and worked on the ethical and responsible use of AI.
After I swear in our panel of witnesses, each witness will
have 5 minutes to provide your opening statement. I know that
will be difficult, but please restrain yourselves. We will then
proceed and have each Senator question for 5 minutes, and we'll
do our best to live within those limits. I hope we will have
time for a second round of questioning. It really depends on
how many of our colleagues join, so could you all please stand
to be sworn in? Please raise your right hand.
[Witnesses are sworn in.]
Chair Coons. Thank you. Mr. Salsberg, you may proceed with
your opening statement.
STATEMENT OF COREY SALSBERG, VICE PRESIDENT AND
GLOBAL HEAD OF IP AFFAIRS, NOVARTIS, WASHINGTON, DC
Mr. Salsberg. Thank you, Chairman Coons, Ranking Member
Tillis, Senators Hirono and Blackburn, and distinguished
Members of the Subcommittee. My name is Corey Salsberg, and I
am vice president and global head of IP affairs for Novartis.
On behalf of the company, thank you very much for the
opportunity to testify today. Today's hearing is timely and
extremely important because, as everyone in the room knows, AI
is already here.
It's in wide use across industries and society, and it is
fast shaping our future. As an early adopter of AI in the life
sciences, Novartis is using today's AI tools to help us
accelerate drug discovery, to make our clinical trials more
efficient, and even as an aid in molecule design. As some
examples, machine learning tools are helping us more quickly
screen the millions of compounds in our libraries to accelerate
the selection of candidates with desirable properties, such as
the ability to bind to a disease-implicated target in the body.
Our AI-enabled platform Nerve Live uses AI tools to help us
anticipate and resolve operational issues at our thousands of
global clinical trial sites, and in a field known as generative
chemistry our scientists recently trained, guided, and used our
AI-enabled research platform called Jaeger to assist in
generating new virtual molecules for potential use in treating
malaria.
Tomorrow's AI tools promise to expand on these activities
and enable many more, helping us to unlock medical insights
like the secrets of the human genome and to ultimately develop
more effective, personalized, safe, and efficient medicines.
What this means is that the time is right to ensure that the
right laws and policies are in place to enable that future. But
the right policies depend on the right information being shared
and the right questions being asked, which makes this hearing
especially timely.
While everybody knows that AI is here, not everyone
understands or agrees what that means. It's easy and perhaps
tempting to equate the outputs of a generative AI chatbot or
the familiar sounds of an AI-generated pop song with human-
level thoughts and creativity and equally easy to assume that
the role that AI plays in one field or situation is the same as
in all others. But when it comes to AI, ubiquity does not mean
uniformity.
Last month, with reference to our Jaeger platform that I
mentioned, a headline in Politico posed the question, ``Can
Jarvis hold a patent?,'' apparently referring to the fictional
artificial superintelligence entity Jarvis in the Marvel movie
franchise. While this may make an intriguing headline and it
certainly delighted my three kids, Jaeger is not Jarvis, and
whether AI can or should be awarded a patent is not, in our
view, the right question at this point in time.
As a foundational matter, before addressing whether AI can
earn or own a patent, the right question is whether today's AI
is even inventing. In our experience, as a matter of fact, the
answer is no. At least in our field, AI tools today are still
just that, tools that are helping to facilitate, optimize, and
enhance human activity and ingenuity and to advance human-
defined goals.
Non-generative machine learning tools that improve or
accelerate tasks we would otherwise do ourselves or that help
reveal information that we're unable to observe on our own are
largely analogous to other tools, that such as high-throughput
screeners, calculators, and microscopes. While generative AI
tools like Jaeger may be used to generate new things,
generating is not the same as inventing, at least not where it
takes a human to define the problem, guide the solution, and
recognize, appreciate, and synthesize the results. But that
does not fully resolve the matter.
A second question, which is top of mind for innovators like
us, is whether using AI to assist in R&D compromises the
ability of our human scientists to patent the resulting
inventions. Here, as a matter of law, we believe the answer is
no. But we must ensure that the doctrine of inventorship law,
known as conception, which traditionally considers whether an
invention occurred inside the mind of the inventor, is not
applied too rigidly to deprive human inventors of patent rights
simply because generative AI now allows some of the inventive
process to occur in silico or on a computer.
Other principles of existing competition law such as the
ability--conception law, sorry, such as the ability of an
inventor to obtain ideas and materials from others so long as
he or she maintains intellectual domination over the process
and the rule that an invention is only conceived when it is
operable, recognized, and appreciated, can address this concern
provided the Patent Office and courts properly apply them.
Because conception law is unique to the United States, though,
we believe this is a critical issue to get right, so we do not
put our Nation's innovators at a disadvantage and compromise
America's economic and strategic leadership.
A third important question is whether we will even need
patents if and as AI advances to the point where it's able to
fully invent on its own. Here, as a matter of policy, we
believe the answer is resoundingly yes because the patent
system's constitutional goal of promoting progress cannot be
realized unless innovators are encouraged to not only invent,
but to publish, develop, and commercialize their inventions. In
this regard, it's important to remember that in our field, the
invention of new molecules, proteins, and other substances is
only the start of the long, complex, and risky process of
developing new medicines.
Without patents or comparable incentives to enable that
work, we would not have new treatments and cures, no matter how
many new molecules appear on computer screens. Thank you very
much for the opportunity. I look forward to your questions.
[The prepared statement of Mr. Salsberg appears as a
submission for the record.]
Chair Coons. Thank you very much, Mr. Salsberg. Ms.
Sheridan.
STATEMENT OF LAURA SHERIDAN, HEAD OF PATENT POLICY,
GOOGLE, NEW YORK, NEW YORK
Ms. Sheridan. Good afternoon, Chairman Coons, Ranking
Member Tillis, and Members of the Committee. Thank you for the
opportunity to appear before you today. My name is Laura
Sheridan and I'm head of patent policy at Google. For more than
20 years, I have witnessed firsthand how patent policy and the
law have adapted to match the pace of innovation. Today, I look
forward to answering the Committee's questions on artificial
intelligence and patenting.
Google's approach to artificial intelligence is both bold
and responsible. We believe we must develop AI in a way that
maximizes the positive benefits to society while addressing the
challenges. The only way to be truly bold in the long term is
to be responsible from the start. We're committed to developing
AI technology responsibly, applying strong safety and security
practices, and continuing to develop our AI innovations in
accordance with best practices.
We introduced our AI principles in 2018 to guide our
development of this critical technology, and we work closely
with external partners to draw on their expertise and share our
learnings on responsible AI practices. Google is utilizing AI
across our entire business. We use AI in our groundbreaking
products used by people everywhere, our contributions to
scientific advances that benefit people, and our efforts
helping to address societal challenges.
The potential of AI to solve big problems is rapidly
increasing. We're proud of efforts to partner in the use of
this technology to help address problems and improve the lives
of people around the world. For example, we're deploying AI to
help forecast floods, monitor prenatal health, and detect
genetic variations linked to disease. We're also using AI in
the innovation process to help expedite chip floor planning,
identify optimal neural network architectures, and improve upon
the drug design process. These advances are described in
greater detail in my written testimony.
We have significant experience seeking coverage for AI
inventions in the United States and around the world and have
one of the largest portfolios of AI-related patents. And we are
engaging extensively with the U.S. Patent and Trademark Office
on issues related to inventorship and artificial intelligence,
including by submitting comments, offering testimony, and
participating in listening sessions.
We agree with the USPTO that there is a spectrum of
inventive behavior. We believe current industry uses of AI are
well within the zone where humans are properly named as
inventors and where AI is leveraged as a tool in the invention
process. We expect to remain in this zone for some time.
However, we do encourage the USPTO to shed light on
inventorship, and we believe that guidance from the office
would allow for a clear conversation between patent applicants
and their counsel.
We also respectfully suggest that the USPTO encourage
patent applicants to document inventor contributions. And we
encourage the USPTO to implement robust technical training for
any patent examiner who is examining AI-related innovations,
whether those are inventions on core AI technology or on
specific applications of core AI technology. The number of
patent examiners who are now examining these AI-related
inventions makes up a substantial portion of the examining
core, and that number is only going to increase.
We urge that a comprehensive technical training program be
put in place so patent examiners are well-situated to assess
whether or not to grant a patent. Google also supports
increases to patent fees for large companies, while ensuring
that small and micro entity fees remain low. We believe the
USPTO will be in the best position to issue robust and reliable
patent rights, including in complex emerging technologies like
AI, when the fees before the grant of a patent more closely
match the costs.
We believe large companies like Google should shoulder more
burden and support the overall health of the patent system. We
appreciate this important discussion on the current state of AI
and patenting. Google continues to engage with Government
officials and the public to further discussions on AI and
responsible development principles. Google will continue to
provide feedback on the intersection of AI and patents to
ensure we continue to strike the right balance for our patent
system to incentivize AI innovation.
Thank you for the opportunity to appear before you today,
and I look forward to your questions.
[The prepared statement of Ms. Sheridan appears as a
submission for the record.]
Chair Coons. Thank you, Ms. Sheridan. Professor Abbott.
STATEMENT OF RYAN ABBOTT, PROFESSOR OF LAW AND HEALTH SCIENCES,
UNIVERSITY OF SURREY SCHOOL OF LAW, GUILDFORD, UNITED KINGDOM,
AND ADJUNCT ASSISTANT PROFESSOR, DAVID GEFFEN SCHOOL OF
MEDICINE, UNIVERSITY OF CALIFORNIA, LOS ANGELES, CALIFORNIA
Professor Abbott. Thank you, Chairman Coons, Ranking Member
Tillis, and Members of the Committee. A picture is worth a
thousand words, and I've tried to do one better with a 3D-
printed model of an AI-generated invention. We call this the
fractal container because it's a beverage container based on
fractal geometry. The United Kingdom Intellectual Property
Office, the U.K. analog of the United States Patent and
Trademark Office, conducted a substantive examination on this
and determined it was new, inventive, and useful, which are the
major requirements to get a patent.
The invention came from an AI system called Davis. Davis
was built and trained by people, but on general knowledge
rather than on knowledge specifically about container design.
It was not asked to solve a specific problem, and its user had
no training in container design or specific intent to design a
new container. Davis was built to combine simple ideas and
concepts into more complex ones and identify when a complex
concept had a positive outcome.
Here it identified that a beverage container based on
fractal geometry would have higher surface area, which would
improve heat transfer like the opposite of a thermos, and it
would make it easier to grip. All that before a human being
laid eyes on the AI's output. If the AI had been a human being,
it would be the patent inventor. When I actually had one 3D-
printed, I was surprised at how easy it was to grip, which
might make it easier for someone with Parkinson's disease to
use.
I was also surprised to see how difficult it was to drink
out of, and it is effectively impossible to clean. But it is an
invention rather than a commercial product. And yet
troublingly, the United States has emerged as an international
outlier in terms of refusing to provide a patent for this
invention, not because the invention itself isn't deserving of
protection, but because of how it was invented.
I know that because I have been leading, on a pro bono
basis, a series of legal test cases on the patentability of AI-
generated inventions and the copyrightability of AI-generated
works, the outcome of which has sadly been to see U.S. courts
adopt a strongly antipatent position. We didn't have a
traditional human inventor in our case, so when we filed the
applications, which we've now done in 18 jurisdictions, we
transparently disclosed it was invented by an AI.
Different jurisdictions have very different rules about
inventorship. Some jurisdictions don't require inventors to be
disclosed, others allow companies to be inventors, which is not
so strange. The U.S. has allowed companies to be authors under
the Copyright Act for more than a century. The United States
has a strict test for inventorship. The person who conceived of
an invention, and usually an inventor, has to sign a
declaration under threat of criminal sanctions that they
genuinely believe themselves to be an inventor.
The Federal Circuit last year held for the first time in
Thaler v. Vidal that if you don't have that traditional human
inventor, you can't get a patent. Of course, no one was ever
arguing that the AI would own the patent. Most inventions are
not owned by inventors, but rather by their employers. An AI
isn't a legal person, it can't own property, and I see no good
reason to change the law to allow that. But it is a bedrock
principle of property law that you own property made by your
property.
So if I have a 3D printer make a physical beverage
container, I own that. There is no reason why I should be any
less entitled to intangible property made by my AI. That's the
right outcome, as a matter of policy, that the owner of an AI
system should own inventions made by that system because the
system is designed to provide incentives regardless of how
inventions are made. Allowing this to be patented will provide
encouragement to the use and development of inventive AI that
will generate more socially valuable inventions. It will
encourage the owners of these systems to disclose inventions
rather than keep them as trade secrets. And it will encourage
the investment needed to commercialize inventions.
So what should Congress do? Amend the Patent Act to allow
the protection of AI-generated inventions by prohibiting
patentability from being based on how an invention is made.
Designation of inventorship is less important than
patentability, but it would be best to keep legal inventorship
based on factual inventorship.
In other words, where an AI has invented something, it
should be disclosed that it is AI-generated. This will tell the
public how it is made, it will facilitate ownership
determinations, and it will keep people from claiming credit
for work done by AI. That's not unfair to AI, which has no
self-interest, but it's unfair to other inventors displaying
legitimate ingenuity and by equating their work to someone just
asking an AI to solve a problem.
Protecting AI-generated inventions and AI inventorship is a
simple, common-sense solution that will not cause other
complications.
And it is critical for Congress to act on this now, not
just because of the state of AI today, but because investments
need to be made now in the AI and innovation of tomorrow. And
there is a universal consensus that AI is only going to get
better at doing this sort of thing. If the right frameworks are
in place, the outcome of that should be significant benefits
for everyone. Thank you.
[The prepared statement of Professor Abbott appears as a
submission for the record.]
Chair Coons. Thank you, Professor Abbott. Professor
Villasenor.
STATEMENT OF JOHN VILLASENOR, PROFESSOR OF LAW, ELECTRICAL
ENGINEERING, AND PUBLIC POLICY, AND FOUNDER AND FACULTY
CO-DIRECTOR, INSTITUTE FOR TECHNOLOGY, LAW, AND POLICY,
UNIVERSITY OF CALIFORNIA, LOS ANGELES, CALIFORNIA
Professor Villasenor. Chair Coons, Ranking Member Tillis,
Members of the Subcommittee, thank you for the opportunity to
testify at today's hearing. I'm testifying today on my own
behalf and not on behalf of any institution. When discussing
the intersection of AI and patents, it's helpful to identify
three categories.
First, there can be patents about AI. Second, there can be
patents describing inventions created using AI. And third, AI
can be used to write patent applications or public disclosures
intended to serve as prior art.
Regarding patents about AI, my point is simple. The U.S.
Patent and Trademark Office is well equipped to handle patent
applications for inventions about AI and has been doing so for
years.
Now, the second category, inventions made using AI, poses a
complex set of policy questions. First, should AI inventions be
patentable? Second, how should inventorship be handled?
I believe that AI inventions should be patentable, and that
inventorship should be attributed to the natural persons who
use AI as a tool to enhance their ability to innovate. More
specifically, as I explained in a recent law review publication
in the Santa Clara High Technology Law Journal, conception
should encompass ideas formed through collaboration between a
person and tools that act as extensions of their mind.
To operationalize this approach, I believe that no
statutory changes are needed. I believe that this view of
conception is consistent with current U.S. patent law. Some
people propose that AI systems should be named inventors. I
respectfully disagree. First of all, as we know from the Thaler
litigation, this would require a change to U.S. patent law.
Many Members of Congress would be rightly concerned about the
consequences of that change. There are other problems as well
that would arise if that change were made.
An AI system cannot provide the required inventor's oath or
declaration that must accompany a utility patent application.
An AI system cannot get deposed in litigation regarding an
invention. It can't give sworn testimony.
Another approach that has been proposed is to deem AI
inventions wholly unpatentable. I do not support this approach.
It would disincentivize investment in the use of AI in areas
where it has high potential. For instance, a pharmaceutical
company would be unlikely to make significant investments in
AI-assisted drug development if it expected that any resulting
drugs would be deemed unpatentable. It would create a new
category of patent ineligibility based on having used too much
AI when making an invention. The resulting ``How much is too
much?'' question would generate years of confusion and
litigation.
The third and final category is the use of AI to write
patent applications or preemptive prior art, publications
designed to foreclose patentability. Let me emphasize here that
the use of AI writing tools to help create explanatory text or
figures regarding inventions is not inherently problematic. It
is reasonable for an inventor to use AI as a timesaving tool
for describing an invention.
However, a problem arises when AI is used to describe
alleged inventions for which there is no conception by a human.
Computer algorithms, whether or not AI-enabled, can be used to
write and publish large online disclosure databases intended to
foreclose patentability. This idea is not new. There is at
least one website in this category that has been online for
multiple years.
To the extent that such publications occur without any
substantive nexus to human understanding of their contents,
there is a good argument that they should not count as printed
publications under 35 U.S.C. 102(a). The entire concept of
prior art is tied to what a person of ordinary skill in the art
would know. The person of ordinary skill is a hypothetical
person who is presumed to know all prior art in the relevant
field at the relevant time.
But I do not believe that a person of ordinary skills
knowledge should be so broad as to include purported art
embedded in computer-generated text for which there is no
evidence that any human has ever understood the significance.
In closing, I will emphasize that AI has extraordinary
promise. In the mid-21st century, maintaining U.S. global
economic competitiveness will require that the United States be
a leader in AI innovation. A patent system that incentivizes
the use of AI to innovate is a key component of that
leadership. Thank you, and I look forward to your questions.
[The prepared statement of Professor Villasenor appears as
a submission for the record.]
Chair Coons. Thank you, Professor. Last but not least,
certainly, Ms. Elluru.
STATEMENT OF RAMA ELLURU, SENIOR DIRECTOR, SOCIETY AND
INTELLECTUAL PROPERTY, SPECIAL COMPETITIVE STUDIES
PROJECT, WASHINGTON, DC
Ms. Elluru. Chairman Coons, Ranking Member Tillis, and
Members of the Committee. Thank you for the honor to testify
today. I'm here to discuss the implications of artificial
intelligence on our intellectual property regime. Specifically,
I want to share the IP recommendations from the final report of
the National Security Commission on AI, the independent
commission Congress created to provide recommendations for
advancing the development of AI to further U.S. national
security needs.
The United States is in a global technology competition.
The Chinese Communist Party understands that the way to achieve
dominance is through supremacy in key emerging technologies. AI
is at the center of this competition. I emphasize that AI is a
foundational technology. AI will influence nearly every aspect
of our lives, especially as generative AI becomes ubiquitous.
AI is a technology like electricity, upon which other
technologies are built.
AI in combination with other technologies like faster
semiconductors and biotechnology, will only accelerate our
discoveries. We are in a race to develop the future of AI, a
race we are in with China. This innovation competition will
shape the world's future. The nations that hold the leadership
and dominant market share in the combination of emerging
technologies will be able to reinforce their societies and
their economies, and importantly, assert geopolitical
influence.
Which nations and how these nations shape the development
and adoption of these technologies will determine the
geopolitical order. The United States must lead in the tech
competition because at stake is a values competition. We want
these technologies to be developed and used according to our
standards and norms, which is the opposite of how China is
using this innovation, such as surveillance of its citizens and
oppression of minority groups. And we cannot do it alone. The
United States, along with our many close partners and allies,
must maintain our technology lead over China.
So how does the United States retain its global tech
leadership? We leverage the strength that has allowed the
United States to outpace and outmatch every technology
challenger we have faced over the past century. Our strength
and power is American innovation. Intellectual property rights
have historically been a critical lever in ensuring America's
innovation leadership.
America's IP regime has spurred American ingenuity since
the 18th century with the 1790 Patent Act. The last major
overhaul of the patent system was in the early 1950s, right
before the field of AI emerged in 1956. Patents are property
rights that incentivize new ideas and inventions by rewarding
inventors for sharing valuable information with the public
domain. These incentives are especially important now when the
majority of funding and tech development in the United States
is from the private sector.
The NSCAI addressed the asymmetry in U.S. and PRC IP
strategies in Chapter 12 of its final report. What can the
United States do in practical terms to implement patent
policies that promote AI innovation?
First, the United States must recognize IP policy as a
national priority and require the development of a
comprehensive plan to reform and create IP policies that
further our technology competitiveness goals.
Second, we need an entity with IP expertise that can
coordinate the various IP equities across our Government to
develop and propose IP policies on an ongoing basis.
Third, the United States must elevate and coordinate
technology policy in the White House by empowering a single
entity to implement a comprehensive technology strategy, a
strategy that integrates IP proposals.
Last, we must prioritize IP considerations that need to be
assessed. The NSCAI proposed a non-exhaustive list of 10 IP
considerations, which are included in my written testimony.
Last, at the Special Competitive Studies Project, we are
exploring a number of these considerations in partnership with
the Renewing American Innovation Project at CSIS, which is
headed by US--former USPTO Director Andrei Iancu. I'm happy to
go into any of these topics during our discussion. Thank you
again for the opportunity to be here today.
[The prepared statement of Ms. Elluru appears as a
submission for the record.]
Chair Coons. Thank you, Ms. Elluru. We'll now begin a 5-
minute round of questions for the members of the panel who are
here--Members of the Committee to ask members of the panel.
If I might, Mr. Salsberg, just to open, could you speak a
little bit more about the difference between non-generative AI
and generative AI in the biotech field and why generative AI
may raise patent law issues that other AI doesn't? In the
example you gave of developing new molecules that might be
useful in the fight against malaria, you referenced Jaeger.
Why is Jaeger not an inventor in the example you gave when
it was generating novel virtual molecules? And why are the
human scientists the inventors, and the AI simply a tool?
Mr. Salsberg. Thank you for the question, Senator. So all
AI today, at least at Novartis, and I cannot speak for other
fields, are tools that are trained on data to basically
recognize patterns and apply them in new ways, either to
existing data in the case of non-generative AI. So it's
learning a pattern, then looking at our data, such as our
compound library, or data from past clinical trials, or images
of cells after different experiments, and trying to find a
similar pattern there. That we call non-generative because, as
the name implies, nothing new.
Generative AI is basically doing the same thing. It is
being taught a pattern. It's being taught to recognize it, but
it's applying that pattern then to create something new or
generate, I should say is a better word, something new. In the
case of Jaeger, I think the reason why this raises concerns is
because of the confusion between generation on the one hand and
invention on the other.
We would say they're not the same because Jaeger, at least
in our example, is not an inventor, because both factually and
legally, what it's doing is not equivalent to human
inventorship. It neither identified a problem, at least in our
case. In our case, the problem is malaria. Right? It was to
design new malaria drugs. It didn't know that there was malaria
out there. It was told, do a drug that does this. It didn't
consider how to address the problem.
It had to be seeded with 20--or taught first with 20,000
examples of compounds, then seeded with three lead compounds
that we know have antimalarial properties to teach it how to
design a molecule. After that, the key point, I think, between
invention and non-invention is an area of the law called--its
recognition and appreciation. That is part and parcel of
conception, and also just factually from common sense, we know
that to be conscious, we have to recognize what's going on.
We don't believe that at this point in time, even
generative AI is recognizing anything that it is doing. Last
point I'll make really quick. I know I'm short on time for your
questions, but a lot has to be done after these molecules
appear on a computer screen. They need to be validated. They
need to be synthesized, which is actually not obvious or easy
to do.
Chair Coons. Understood.
Mr. Salsberg. Yes.
Chair Coons. Thank you. Ms. Sheridan, can you imagine a
time in the relatively near future when AI will advance such
that it really is possible for AI to generate meaningful
inventions with no human intervention at all?
Ms. Sheridan. Thank you for the question, Senator. It's in
our written testimony that our current experience and what we
expect to be the experience for, you know, the foreseeable
future is no. The human involvement is so great at this point
when AI is being used as a tool in the innovation process from
start to finish.
Chair Coons. You mentioned guidance from the PTO and
inventorship would be really helpful. You expounded on that
just a little bit. Is there anything you missed that you think
we should hear about what that guidance from the Patent and
Trademark Office should look like as we move forward in
examining this intersection between patent law and AI?
Ms. Sheridan. I would have it really look at two big
things. One, it should be broader than AI. I think AI is
creating this conversation, but I think inventorship is
challenging. So I think that guidance should more broadly
encompass AI usage, but also other tools. You know there are
computer simulations. There are other things that are being
used in the innovation process.
And I think, too, it really should give comfort to
inventors that they are doing their best to memorialize the
inventorship for their patent applications and, you know,
taking a position from the PTO in terms of what we're talking
about with joint inventorship, that the humans are properly
named.
Chair Coons. Professor Abbott, you mentioned--you've led or
managed I think it's 18 now foreign patent cases involving Dr.
Thaler's AI-generated inventions. And you called the United
States an outlier in failing to recognize inventions that are
AI-generated for patent protection. What are the consequences
of that? What do you see as the costs or the risks to the U.S.
system? And you asserted that it wouldn't cause any other
complications for us to modify patent law in a way that allows
essentially co-invention.
I'd be interested in whether other members of the panel
agree, and if you'd address my question about what are the
consequences of our being an outlier.
Professor Abbott. Sure. Well, I mean, I think the most
pressing commercial problem here is not where you have co-
invention, although that can sometimes be an issue, but where
you don't have a traditional human inventor you could identify.
And I agree, in Mr. Salsberg's example, you have that or a good
argument for it, where you do have a lot of human intervention
and there's always human intervention. The question is, under
inventorship law, there's a thing that someone does that makes
them an inventor.
And sometimes that thing is now being done by a machine,
and perhaps not that much right now, but it is increasingly
going to be the case that AI is doing this, that companies like
Microsoft are pretraining these drug-seeking AIs on how to
generate new molecules and how to validate them. You can have
AI that will evaluate the utility of its own output.
When you are, for example, modeling industrial components
and you know, I want an industrial component that is lighter or
that is longer or that meets some preset criteria that's well
understood in the art. So as AI----
Chair Coons. So in your instance, was it simply a failure
to instruct that the container should be easily cleanable?
[Laughter.]
Professor Abbott. Well, if someone was now making--you
know, one can get a lot of patents on a single product. Right?
So it might be that this was the first product. Someone uses
that and says, ``You know what? This is great, but I need it to
also be cleanable, so how can we redesign it for that?'' And if
a person did that, they might be an inventor on that second
patent. You know, but basically, it's all a question of degree
and what specific input a person has into invention, but the--
--
Chair Coons. What are the consequences of our being an
outlier?
Professor Abbott. The consequences are that companies will
be discouraged to use AI in the invention process because it
will risk them getting patents on their inventions. Right? And
that will depend very much on the field in which AI is being
used, the systems being used, and the specific facts of how
this is done.
Chair Coons. Thank you. I have a lot more questions for the
panel. But respecting that I'm already over time, I'll yield to
my Ranking Member.
Senator Tillis. I think I would have started by asking AI
how I could develop an accessory that I could pay for--that I
could sell, after I sell the bottle, that cleans it very well.
[Laughter.]
Senator Tillis. I'm sure somebody's probably already
working on that. Ms. Elluru, you--did I pronounce that right?
Ms. Elluru. Correct.
Senator Tillis. Tell me a little bit about your current
assessment of--you and your opening comments were talking about
really Western world versus China in this competition. Where
are we now in terms of the pacing threat of China surpassing
us, or have they?
Ms. Elluru. I think the U.S. is in the lead with respect to
AI, but barely. Had the U.S. as leading in cutting-edge chip
design, but manufacturing of semiconductors is still a
contested space with Taiwan at the center. I think the U.S. is
forefront in novel algorithms and architecture, and it has the
frontier large language models. But China is leading in data in
part due to its pervasive domestic governance and collection
regime.
And I think with respect to large language models, it is a
little bit hindered because these large language models are
trained on the internet, and the internet is mostly in English.
They also have censorship, which constrains their access to
data.
Senator Tillis. You also mentioned in your opening
testimony--you didn't use these words, but I inferred from it
that you feel like the administration needs a kind of whole-of-
Government view with respect to AI? But would you elaborate on
that? What would that look like?
Ms. Elluru. I think, as we recommended at the NSCAI, that
we need a single entity with IP expertise that can coordinate,
and importantly, unify the different IP equities across the
Government. And we recommended that that sit with the United
States PTO Director as well as the Department of Commerce
Secretary, and that other Federal Government agencies and
departments resource efforts and--resource their efforts to
support that, this PTO Director.
Senator Tillis. Mr. Salsberg, you mentioned in your
testimony clarifying the law surrounding conception would
restore certainty. Can you state that if too narrowly
construed, the law could potentially be misapplied to deprive
human inventors of legitimate patent rights to their own
inventions?
Mr. Salsberg. Yes, we do believe that if it's too rigidly
applied, that could be the result, because we have a standard
that has been established in a different context. It was about
priority of inventorship, meaning which of two inventors
invented first. For 100 years, we've had a law developed that
looks at what happens in the mind of the inventor. And we
believe that that is way too rigid of a standard.
But we also think that existing law has multiple principles
available. I know to be very brief in my answer, but there's a
case called Morse v. Porter which says that as long as an
inventor has intellectual domination over the process, you can
even adopt key ideas from outside. And also, if you don't
recognize, appreciate, and have an operable invention at the
time that it is created, it is not actually conceived until
someone else recognizes and appreciates it.
So, we think combining these doctrines together--and
there's a bunch of cases in our written testimony that stand
for these propositions--if the courts and Patent Office apply
this properly, we can resolve that issue under existing law.
Senator Tillis. Just briefly, could you explain how we
compare to other countries, other jurisdictions?
Mr. Salsberg. Yes. We're the only country in the world, to
the best of my knowledge, that has anything like a conception
requirement. In much of the world, you don't even need to--you
never even get asked about inventorship. Many countries, you
list an inventor and it's just of your own accord. It only
comes up in situations where a remuneration dispute or a joint
inventorship dispute comes up. Some countries don't even
require you to list an inventor, like Israel and Austria.
Senator Tillis. Final question, I may also stick around and
ask some further--not a question, but you know, it feels like
to me, in everything that we do on Capitol Hill, we're doing
well when we're providing certainty in various areas. And I
feel like if we don't tackle these IP issues, that we are
likely to be the jurisdiction where these inventions occur.
Would anybody disagree with that? I mean, there is work that we
have to do.
There are some that are saying it can work within the
current law, but I, for one, feel like we have to have
certainty and clarity, or we really run the risk of having the
gap. Maybe our competitive advantage today will not be the same
just a few short years from now, given the pace, the rate of
change that we're observing in AI space. Anybody disagree with
that? That it's not broke, don't fix it. We've got something
we've got to work on here. Professor Villasenor?
Professor Villasenor. As I've argued in the law review
article I mentioned, and in the testimony, I think U.S. patent
law already accommodates AI inventions and that the inventor
should be the human who uses the AI as a tool, as an extension
of their mind. I think that's consistent with the understanding
of conception that's been in place since--and actually
conception is not defined in Title 35. It actually comes from
an 1890 treatise by Robinson, and the Federal Circuit has cited
that as recently as, I believe, 2021.
And I think just viewing the conception through that
broader lens, consistent with that original, commonly cited
definition, is enough to encompass and take account of
inventions made using AI.
Senator Tillis. Yes, I think my concern is, if we don't
identify some of the bigger threats here and provide clarity,
this could be sorted out in the courts. Which is one of the
reasons why we're working on patent reform because we have a
lot of really bad decisions that came out of the courts with
people who really didn't understand the underlying concepts of
the case they were overseeing. Thank you, Mr. Chair.
Chair Coons. Thank you, Senator Tillis. Senator Hirono.
Senator Hirono. Thank you, Mr. Chairman. So, the thing with
AI is that it is being developed, the use of AI, through so
many--in so many different ways is happening by leaps and
bounds. And as we sit here struggling to figure out a way to
regulate AI in an appropriate way without really discouraging
invention, etc., it's not the easiest thing to try to decide
what to do. And in fact, none of you mentioned the, for
example, the White House blueprint for an AI Bill of Rights and
the Bill of Rights that individuals should have.
And as we sit here knowing that some entity out there can
replicate our voices, put out a campaign ad that we've had
nothing to do with, and what are we supposed to do about that?
We have no awareness of it. We don't even know if this was an
AI-generated thing. So, the implications are vast on how we
should be regulating AI. We just know we should, can't decide
what to do quite yet. But it's interesting to hear your
testimony about what should be patentable and after the Thaler
decision, where it seems very clear that the human element has
to be there in order for something to be patented.
Except, I believe, Mr. Abbott, you know, your approach
seems to be so simple. Whoever owns the AI system should own
whatever is invented through that system. I'd like to know from
the rest of the panelists whether that is the way we should go,
in spite of the Thaler decision. Let's start with you, Mr.
Salsberg.
Mr. Salsberg. Thank you for the question, Senator. That is
one option. If and when we get to the point where AI is truly
inventing. As I said in my testimony, we're not sure we're
there yet, but if and when we get there, we totally----
Senator Hirono. Well, things are moving so fast that I can
envision that we're going to get there. So, go ahead.
Mr. Salsberg. And when we get there, we're in full
agreement that we do need to make sure that patents still issue
to those who use AI to create useful inventions for society and
promoting progress. We think there are three ways to address
it, though. One is to award a patent to AI. Another is to
simply change our inventorship standards and expand what it
means to invent. And we believe, again, we can get there by
taking existing principles of conception law and applying them
properly.
And if that doesn't work, we also think we can eliminate
the inventorship requirement in most situations and get to the
same result without having impact on other aspects of patent
law, such as the level of ordinary skill in the art of an AI
and the inventor's oath and things like that.
Senator Hirono. Ms. Sheridan.
Ms. Sheridan. Thank you for the question. You know I agree
with Mr. Salsberg. Our current experience is the human is very
much involved. The human is the inventor. I have kind of two
concerns with the approach of going with, you know, the system
being an inventor or the ownership vesting in the owner of the
system. I think one just sort of sits with the Constitution and
the--really the reason for the patent system, which is to
incentivize the behavior of humans to innovate.
And you know, I question whether that incentive is needed
for an AI system. I think that's not really the purpose of the
patent system in the first place. So I would want to think more
about, does that line up with the purpose of the system.
I think the other question, though, which is more the one
that would keep me up at night is obviousness. You know,
inventions need to be nonobvious, and if you have this ability
for an AI to be named, what does that mean for the obviousness
test? Where is the bar? And who are you kind of measuring this
against? And how does a patent examiner, one of 8,000 in the
PTO's workforce, apply that?
So, I think there are a lot of open questions that would
come up from what seems to be a simple fix.
Senator Hirono. I'm going to move on to Mr. Villasenor.
Professor Villasenor. Yes, very briefly, I'll just say that
everything an AI system does is ultimately traceable back to
one or more people. And I think that when we focus only on the
AI system, we miss potentially the incredibly important and
relevant in this context, contributions of the people who have
ultimately enabled an AI system to do what it does.
Senator Hirono. So you're not particularly troubled by the
idea that whoever owns the system should own whatever the
system generates?
Professor Villasenor. Well, actually, I am in the sense--in
the sense that I think the owner--I mean, whoever owns the
system shouldn't be the dispositive question. Again, I believe
that the people who use AI as extensions of their minds should
be the inventors. And that would be true even if they, for
example, you know, leased under a contract, an AI system that
they use to expand their creative capacities.
Senator Hirono. Mr. Chairman, could I ask Ms. Elluru, too?
Chair Coons. Absolutely. You got it.
Ms. Elluru. Thank you, Senator. I have two comments to
make, and the first one is open source. If we recognize patent
owners as having the right, what happens when a user uses open
source models to invent? And the second comment is, with
respect to obviousness, I think that AI is going to be commonly
used in the invention-creation process. So, I think we will
have the obviousness issue regardless of whether we recognize
AI as an inventor or not.
Senator Hirono. Thank you, Mr. Chairman.
Chair Coons. Thank you, Senator Hirono. Senator Blackburn.
Senator Blackburn. Thank you, Mr. Chairman. Thank you to
each of you for being here and for your testimony. And I so
appreciate that we are looking at this issue. I'm from
Tennessee. We have a lot of entertainers. The voice cloning is
an issue. ChatGPT, being able to write a song that sounds like
it would be in the style of so-and-so, some artist--that's an
issue. And making certain that we're able to preserve these
intellectual property rights is--this is coming up a lot as we
talk with our entertainment community.
And then also, we have a lot of people that are doing work
in the energy space, automobile manufacturers, and they're
busy. Ford has done a huge investment in Tennessee with SK
batteries, billions of dollars there. GM has its Ultium battery
facility in Tennessee. Now, these innovators are saying, ``I
want to hold my invention.''
And of course, China, which we've discussed some today, is
a pro at lifting intellectual property and not protecting that
intellectual property. And I look at what they're doing,
stealing it, our intellectual property, the emergence of China,
and their aggressiveness in the field of AI. And Ms. Elluru, I
want to come to you on this. I pulled a number, 2021. China was
the global leader for patent filings, and they had 1.59 million
filings. They were double us. Then 2011 to 2021, Chinese AI-
related patents accounted for nearly 75 percent of the global
total. That should be a wake-up call to every one of us,
especially those of us that represent so many innovators.
And I decided to go on and look at what was happening with
PTAB because so many of our innovators have their patents
challenged. And China has been very aggressive using this. And
primarily ZTE and Huawei are the companies that are the
leaders, if you will, in challenging U.S. patents.
So, Ms. Elluru, as we look at this and we look at the
global marketplace, I want you to go a little bit further than
you did in your response with Senator Tillis and talk for a
minute about that threat that is there with China trying to
dominate in this space to the point that they've doubled our
number of filings.
They've got 75 percent of that global share, and then
they're coming at it from the other angle, trying to discredit
our patents.
Ms. Elluru. Thank you, Senator, for that question. I think
with respect to China's prolific patent filing, quantity does
not always equate with quality. However, there are implications
with those numbers. First, they're using it as a narrative that
they're winning the tech competition.
Senator Blackburn. Well, and that they're kind of blocking
competitors. Right?
Ms. Elluru. Absolutely. And so I think there's a real
parties in interest issue, too, with China abusing--using our
system against us.
Senator Blackburn. I agree with that. And I think it's why
we have got to push back on that narrative, because of the
numbers and the way they're trying to be the first to file in
certain areas.
Ms. Elluru. Agreed.
Senator Blackburn. Yes, I have quite a few concerns about
that. And what they're going to do with really pushing through
on this system and using the system of protection we have to
advantage themselves. I do want to move to one other thing. I'd
like for you to speak for just a minute about what we do to
counter this aggressiveness. As you look at it, what steps
would you say we need to take? And anybody else after Ms.
Elluru weighs in on this, anybody else that has a suggestion,
I'd love to hear that.
Ms. Elluru. Thank you, Senator. I think the first thing we
need to do is recognize IP policy as a national security
priority across our departments and agencies. And as I
mentioned earlier, we need a central entity----
Senator Blackburn. Yes. I heard you say.
Ms. Elluru [continuing]. With IP expertise that can
coordinate across the Government.
Senator Blackburn. Okay. Anyone else? Anything to add?
Mr. Salsberg. Yes. Thank you, Senator.
Senator Blackburn. Mr. Salsberg.
Mr. Salsberg. I would just quickly add that we should fix
patent eligibility under Section 101 for both software and
diagnostics.
Senator Blackburn. Okay.
Mr. Salsberg. And two, we should clarify conception law, if
the courts--and this body should do it if the courts in the
Patent Office don't apply it properly to make sure that the use
of AI is encouraged and adopted by American innovators.
Senator Blackburn. Okay. Anything else from any of you? No.
Thank you, Mr. Chairman.
Chair Coons. Thank you, Senator Blackburn. Senator Padilla.
Senator Padilla. Let me begin by just thanking you, Mr.
Chair, for holding this hearing and a special welcome to the
majority of the witness panel that are Californians. Not a
coincidence. Certainly a timely topic, since our innovation
ecosystem should be a prominent element of our ongoing AI
development and regulatory discussions taking place here in
Congress and within the administration. And our patent laws
obviously play a significant and vital role in shaping that
ecosystem.
I know there's growing anxiety about AI systems and tools'
potential to replace or displace the work being done by humans.
And the questions raised today about how much invention needs
to come from humans to preserve intellectual property rights
mirrors actually some of the discussion taking place in the
world of copyright law. So, I want to take the time--take time
right now to ask about how AI is impacting human invention and
innovation, and I want, to help us focus that discussion, cite
Stanford's AI Index 2023 Annual Report. Quote, ``AI models are
starting to rapidly accelerate scientific progress and in 2022
were used to aid hydrogen fusion, improve the efficiency of
matrix manipulation, and generate new antibodies.'' So, clearly
AI has been, and should continue to be, a great tool for
invention and discovery.
I know I've said a lot, that's just to tee up the following
question and I'll address it to Ms. Sheridan. Can you speak to
the way AI is helping not just your company, but the industry,
invent? And under current U.S. patent law, as you understand
it, how does the use of AI affect the ability of your
respective industries to patent those inventions?
Ms. Sheridan. Thank you for the question. So, I mean, in
our observations and use of AI, it's really been a powerful
tool in the inventing process. We create machine learning
models and then we're able to use those models in further
development, whether it's for chip floor planning or neural
network architecture search. So it's been an incredibly
important tool to us. I think there's two topics here.
One is the use of AI as a tool for innovation, and then
there is simply the topic of AI innovation in general and
making sure that the system is both balanced to encourage both
of those things. And the conversation we're having around
inventorship that really relates to this question of using AI
as a tool in the innovation process and the conversation we
have just around the general patent system and making sure it's
striking the right balance, that's for innovation writ large.
And I think I--there's a statistic from an IEEE report that
shows, you know, 70 percent of all AI patents are held by U.S.
companies as of 2018. And I think that really shows right now
in the U.S. innovation in AI is growing, it is flourishing. And
we have patenting activity that reflects that, both in the core
AI innovations, but also in the use of AI as a tool to
innovate.
Senator Padilla. Thank you. Clearly much to continue
discuss on that point, but also want to make sure I talk about
the innovation trends that we're seeing. AI innovation has
grown at a rapid pace. We all acknowledge that. And the use of
AI by innovators, as Ms. Sheridan just referenced, whether it's
individuals to startups to large enterprises, is also growing
rapidly.
One study found that AI will potentially contribute $15.7
trillion to the global economy by 2030. But the Patent Office
has found that the volume and the share of public U.S. AI
inventions has grown significantly in the last decade alone,
more than 15 percent of all published patent applications in
2020 were directed to AI technologies, equivalent to almost
80,000 patent applications, just to give you a flavor for the
volume here. And there's been a tremendous investment,
obviously, in artificial intelligence R&D among U.S. companies.
Question for Ms. Elluru and Ms. Sheridan: Do you believe
our current innovation ecosystem, including our patent laws,
are incentivizing AI-related innovation? Or how else would we
interpret the increase in AI patent applications?
Ms. Elluru. Thank you, Senator, for that question. I think
there's a lot of uncertainty in our patent laws currently with
respect to patent eligibility, also with respect to whether AI-
generated inventions are patentable. I think that the common
view right now is that AI is a tool, a very sophisticated tool
that's used in the invention-creation process. I think some
people would say that you still need an inventor to frame the
problem set, analyze the output of the AI, recognize and
appreciate the results.
But I think an argument can also be made, especially in
litigation, that the AI is contributing to the conception
process itself. It's augmenting or supplanting cognitive
capabilities. And under current law, an AI machine cannot be an
inventor. So the argument would be that the AI machine was a
co-inventor, but it cannot be an inventor. So under current
laws, that's unpatentable. And I think that uncertainty leaves
to unpredictability, and it diminishes the patent system. Thank
you.
Senator Padilla. Ms. Sheridan.
Ms. Sheridan. Thank you for the question. We have seen AI
innovation flourishing, and I think the patent numbers are one
thing that reflects that. There are other ways we see it in the
investment. You mentioned the growth of the economy, but it's
really important, and I mentioned this in my opening statement,
that training is in place for examiners because you mentioned a
statistic about the growth in AI patenting at the Office. That
means, at some point, the majority of patent examiners are
going to get an invention that crosses their desk that relates
to AI in some way.
And they may not be experts in AI. They're experts in the
field that AI is being applied. And it is critically important
to make sure that examiner knows what they're looking at and
knows, yes, this is a deserving patent, I can grant this, or
no, this is an obviousness issue, or whatever it might be.
Training is really just a fundamental issue that to continue to
incentivize this level of innovation, which we already have, we
need to make sure that PTO examiners are given the tools that
they need.
Senator Padilla. Thank you. And last, just a comment to
Professor Abbott. Please don't show that cup to my kids,
because no doubt they're going to want to put their fruit
smoothies in there, and then I'm going to be stuck scrubbing it
out. Thank you, Mr. Chair.
Professor Abbott. Well, Senator, very briefly, on behalf of
the California academic contingent, thank you for having us and
for not inviting anyone from USC.
[Laughter.]
Chair Coons. Thank you. Thank you, Senator Padilla. Senator
Blumenthal.
Senator Blumenthal. Not going there. But I am going to make
a very non-scientific observation here, at great risk, I must
say, because it could lead to unwarranted conclusions. But I
would say, looking at the audience in this room and the group
waiting outside to get into the room, that the average age of
our audience is roughly one-half the average age of United
States Senators.
Senator Tillis. Senator Blumenthal, we confirmed that. I
had a show of hands.
[Laughter.]
Senator Tillis. There's only one in the audience who's over
the age of 35.
Senator Blumenthal. Which is enormously exciting for me.
And I make this statement not out of any ageism, but I
guarantee that in China or in many of the other countries that
are potential competitors, there are not groups like this one
voluntarily--voluntarily pursuing research, innovation,
invention in this area and coming all the way from other parts
of the country to listen to this kind of proceeding. So, I want
to thank you all for being here today.
And when I thank you, I'm thanking literally thousands,
tens of thousands, hundreds of thousands of young people who
are studying this area and who will be the source of invention.
And you're right, Ms. Sheridan, AI invention and innovation are
flourishing in this country, which should be a source of pride
and excitement to my colleagues and myself. Which brings me to
my first question.
A number of experts in the industry have suggested we ought
to have some kind of pause on innovation or on the release of
new products, put aside the issue of testing them. Even if I
thought it were a good idea, I don't. I think the idea of a
pause is totally impractical. So, let me ask any of you whether
you think a pause on AI--I know it's a somewhat simplistic
question, is a good idea and even doable?
Professor Abbott. Well, just to briefly address that,
although I think you addressed it yourself a bit. I think the
biggest reason a pause would be bad, as Senator Tillis
mentioned, is a lot of social good is going to come out of AI,
and with the right frameworks in place, predominantly good
rather than risk. And pausing it means that we're not going to
get this innovation and creativity and value from AI that we
otherwise get. On top of which, I don't think you really could
enforce a pause. And on top of that, other jurisdictions are
not going to similarly pause so----
Senator Blumenthal. That's my next--unless anybody
disagrees, just in the interest of time, I'm going to move to
my next question. Yes, sir?
Professor Villasenor. Well, I was just going to say I agree
that it would be a bad idea for many of the same reasons,
totally unenforceable, and I could list lots of reasons why
it's a bad idea. I don't support the idea of a pause.
Senator Blumenthal. And the ones who would love it the most
are the Chinese, our competitors in other jurisdictions, as
you've said, Mr. Abbott. So it's a bad idea for a lot of
reasons, and it's impractical and unenforceable. I want to ask
a question because you mentioned the word framework, Mr.
Abbott.
And in a hearing that Senator Hawley and I had in a
different Subcommittee of the Judiciary Committee, the Privacy
and Tech Committee, we focused a lot on this idea of a new
agency, licensing and registration, guardrails and safeguards,
like, maybe what I call the nutrition label, means of
distinguishing and stopping deep fakes, audio clones, and so
forth. Some of the bad outcomes of AI innovation. I have in
mind a framework, probably a new agency that would not in any
way impinge on the current patent system.
Do you think that kind of new agency is compatible with our
current intellectual property protection system? I'm going to
open it to anyone who wants to respond.
Ms. Elluru. Thank you for that question, Senator. I think
that we need to use existing regulators to adopt AI regulation
with existing authorities as in the short and medium term, but
I think also exploring the new agency--a new AI agency and its
functions is a great idea. I think it'll take time. So I don't
think those two recommendations are mutually exclusive, and I
think we need to pursue both at the same time.
Professor Villasenor. I'll just say I think it's important
to recognize the protections we already have under broader
existing frameworks. For example, if a company uses an AI
system to make hiring decisions in a manner that discriminates
based on a protected characteristic, that would already be
unlawful under Title VII of the Civil Rights Act of 1964. So we
already have a lot of the frameworks that would apply to AI,
even though they are not AI-specific.
Senator Blumenthal. Mr. Abbott.
Professor Abbott. Just to briefly weigh into that, I mean,
I think you are completely correct that it is the Government's
responsibility at some level to manage some of these risks for
the public and that having an administrative agency dedicated
to that would be helpful. I do think that as a general matter,
a technologically neutral approach to the issue is the right
way to go.
So, for example, if we want to, and we do prohibit
discrimination based on protected characteristics, that there
are laws that apply to companies doing that, whether you have a
human resource manager or an AI system operating to do that
sort of thing. You know, so I think in principle it's a good
idea, but very much depends on the details of how it's
implemented.
Senator Blumenthal. Thank you. Ms. Sheridan or Mr.
Salsberg, do you have observations? I'm going a little bit over
my time.
Mr. Salsberg. Sure, I agree with the other speakers. I
would just add that it's also important that we make sure that
there's public sector and private sector collaboration in
forming this kind of commission, or at least into the input
because we also want to make sure that we're not undermining
innovation. And I would also add just one more point, which is
that there are other agencies, like FDA in the case of our
industry, that are already looking at these issues, and it's
very important that they continue to be a part of this.
One statistic is that over 100 applications to FDA in 2021
have already contained AI or machine learning components. So
they're already looking to these issues. And if we had a new
commission, we'd have to make sure there's coordination between
those agencies.
Ms. Sheridan. Thank you. I'll keep it very briefly. So our
CEO has said AI is too important not to regulate, and it's too
important not to regulate well. I think our view of regulation
as a sectoral approach, is appropriate, and when it comes to
patents, that would mean the Patent Office kind of handles
issues related to patent policy. But we do believe it's--
regulation is important.
Senator Blumenthal. I welcome all of these comments and
this hearing and the one that my Subcommittee did, I regard as
just the beginning. And I hope that we can call on all five of
you and many others to be involved, because we do, to take Mr.
Salsberg's point, want to regulate. I think everyone has agreed
regulation is necessary, maybe through some kind of Federal
agency, call it an administrative agency.
But we also want to continue to encourage innovation and
the startups and the folks who are in this room who are going
to be inventing stuff in their parents' garages, not to
deprecate anybody who's in this room, but, as you know, that is
a well-founded source of innovation in this country. And we
want to make sure that we don't heavy-handedly deprecate or
discourage innovation. Thanks, Mr. Chairman.
Chair Coons. Thank you, Senator Blumenthal. Before I defer
to Senator Tillis to begin a second round, I just want to
follow up on that point, if I could, because one of my concerns
about how we regulate AI is that we may end up favoring well-
funded and established companies and make it dramatically more
difficult for the startup, the innovator, the category breaker,
smaller companies that are the backbone, often of American
innovation, to compete.
I'd be interested in any brief comments from the panel on
how you think we strike that balance. How do we regulate AI
responsibly without hampering innovation? Ms. Elluru, do you
want to begin? And I'll just welcome any brief comments from
members of the panel.
Ms. Elluru. Thank you, Senator. I think we do so very
carefully. One of the things that we've recommended at the
Special Competitive Studies Project is that we focus our
governance efforts on high-consequence use cases. To going to
Senator Tillis' comments earlier, AI can have benefits as well
as harms, and we need to focus our efforts on those benefits
and harms that are going to have the most significant impacts
on societies and individuals. So it is a risk-based approach,
and we need an American model of identifying what high-
consequence use cases look like.
Professor Villasenor. I'll just say that, as we all know,
new market entrants, new companies have been the source of an
incredibly large amount of innovation historically in the
technology space in the United States. And there's no reason to
believe that's going to change at all in the sense of AI.
And given that, it would be a mistake to implement a
regulatory regime that, as the Senator, as you said, favors the
large, well-funded incumbents who have the resources to comply
with very burdensome regulation while disfavoring the new
market entrants who are going to be the future Googles of the
world.
Chair Coons. Thank you. Professor Abbott.
Professor Abbott. I might make a broader point, which is
that I think when confronted by this technological evolution
and how it's challenging current laws, that you should take a
step back and think, you know, what is the purpose of patent
law and what is it that we want to do. Right?
Someone earlier said, well, it's to benefit inventors, you
know, in copyright law, for example, despite the Copyright
Clause of the Constitution, the U.S Supreme Court has held
copyright law is not primarily to benefit authors, it's
primarily to benefit the American public by encouraging the
creation and dissemination of new works. So do we want a patent
law that is going to specifically encourage human behavior and
discourage people automating?
Do we want something that's going to benefit small
companies over more than bigger ones, or do we want something
that's just going to generate more public good and more
innovation? You know, I think the right place is for Congress
to decide what the goals of the system are and then, you know,
implementing it follows from that.
Chair Coons. Yes. The point that Professor Villasenor just
made was historically across technology broadly, if you come up
with a system that disfavors the innovator, the small, the
startup, then you will, as a consequence, have less innovation.
Would you disagree with that as----
Professor Abbott. No, I agree with that. I took his point
also to be, you know, the more regulatory requirements you
place, it tends to be more burdensome on small and medium
enterprises than companies like Novartis and Google that have
large teams of people that can help comply with this sort of
thing. So, you know, that sort of approach, I think, is more
warranted the higher risk use you're looking at.
Chair Coons. Thank you. Ms. Sheridan.
Ms. Sheridan. Yes, you know, to bring this into a patent
context, I think I support something as simple as fee
increases, and I mentioned that in my opening statement. But I
think if you want to make sure that small and micro entities
before the Patent Office have, you know, a robust examination
when they have that AI innovation before an examiner, you need
to make sure the Office has enough fees to do that examination,
and large companies like Google can support that and make sure
that the resources available to the Patent Office are greater
upfront.
Chair Coons. And of course, we need to permanently end the
pernicious practice of fee diversion.
Ms. Sheridan. We would agree that that would certainly be a
good thing to do.
Chair Coons. One of my goals in getting elected to the
Senate, literally. Mr. Salsberg.
Mr. Salsberg. Yes, thank you. I would just echo the
importance of following a fact-based and a risk-based approach
to this to strike the right balance, and also the importance of
involving the different entities within the private sector in
that conversation. I think it's critically important. Last
point I'd make is the risk-based approach is roughly what I
believe the European Commission is following in its proposed
AI--the AI Act. It's not a perfect Act, but that is the general
approach. And I do think it needs to be tailored to the
specific risks identified in each application of AI.
Chair Coons. Thank you. Senator Tillis.
Senator Tillis. I'm just imagining the scene of where
you're on the stump campaigning for the Senate, the applause
line when you talk about ending fee diversion. I'm sure that's
a crowd-pleaser.
[Laughter.]
Senator Blumenthal. They'd probably applaud the diversion.
[Laughter.]
Senator Tillis. But you know, back to Senator Blumenthal,
he looked at the crowd much the way I did, and it's good news
to see young people focused on this subject.
I think one thing we have to do here is educate the Members
of Congress well in advance of making any policies that could
have a significant impact on regulatory burden or regulatory
certainty.
And I am pleased to hear, I think, that the House--the
Speaker, has actually had a course curriculum put together by
an East Coast institution to educate our Members on AI policy
implications--and I think also quantum computing. I'm going to
reach out and see if that's something that we can get here. I'm
actually going to another meeting here on AI with Senator
Warner shortly. And one of the subjects that we're talking
about are the possible results from improperly developed and
secured models. We haven't talked about this here, but that's
another thing that we can't assume that all AI engines are
equal, all of them are fully vetted, all of them are safe.
I mean, that's a completely different--outside of the
jurisdiction of Judiciary Committee, but something we have to
look at and get a light touch, but start recognizing those
outcomes.
Ms. Sheridan, I should have mentioned to you--I mentioned
that I'm in the ChatGPT beta program, but I have also explored
Google's offerings, Microsoft's adaptation of ChatGPT, and the
dozens of other engines that are out there. One of the things
I'm trying to wrap my head around and honestly, it came to me
as I heard you all talking, is explain to me how--I'm in the
group of let all inventors invent.
I don't care how big you are, I don't care what scale you
have. Let's just have a healthy ecosystem with a lot of mom-
and-pop shops and individuals having a level playing field. But
it would just seem to me, as this space evolves, that there
would--in some respects am I wrong in thinking that this could
be a great enabler for the mom-and-pop shop, the people that
don't--I worked at Wang Laboratories up in Boston in my
research and development. We had floors of computer scientists
and engineers working on things.
It seems like this becomes an invaluable resource to a
handful of really smart people without a lot of capital to
accelerate. Am I missing that, or would you all generally agree
with that? Where's the risk to the small inventor that I'm
missing?
Ms. Elluru. As I said in my opening statement, I think that
AI is going to be very ubiquitous in the invention-creation
process and just use--and so I think it will raise the boats
for everyone, and I think it will make mom-and-pop businesses
much more productive and efficient as well.
Professor Villasenor. I'll just say, I think that one of
the really amazing things about AI is it does help democratize
the innovation process.
Senator Tillis. Exactly the word I've used.
Professor Villasenor. Yes. It puts in the hands of a far
greater number of people, including people who wouldn't
necessarily have had the funds to do these things before, an
enormously powerful set of tools to innovate. And I think, you
know, if we play our cards right with the policy side of it, we
as a country will benefit from that.
Senator Tillis. Mr. Abbott.
Professor Abbott. You know, I would agree with that. And
mentioned that Dr. Thaler's business is a small business
located in the Midwest, and you know, we've seen this,
particularly with the generative AI systems and copyright. You
know, now, I can make creative things that I couldn't have made
before if I wasn't a major motion picture. I mean, they're
still pretty bad, but I have these tools and I can make
something with them.
Ms. Sheridan. I agree. And we're seeing how it can be
useful in the patenting process itself in terms of drafting of
an application, you know, how can these tools be used to help
someone, you know, get their ideas down on paper as a starting
point. So that's yet another way it can kind of help assist in
the patenting process.
Mr. Salsberg. I would agree with that as well, with the one
caveat that I do think it's important that you also have human
supervision and human review of the outputs. We've seen report
after report of situations where ChatGPT and the equivalent
were used and the results ended up being inaccurate. So, I
think right now it's a balance until we are more assured that
the outputs are reporting accurate data, even for a small
business, that they're actually getting the right answer that
they're looking for.
Senator Tillis. And that gets into the--whatever models
you're using and whether or not they're secure, whether or not
they're relying on a base of data that produces a safe or sound
result.
Mr. Salsberg. That's right. The quality of the data is
very, very important.
Senator Tillis. Yes. Well, I appreciate--I think this is
going to be one of many hearings that we have on this and maybe
work groups as we move forward if we can find any areas for
legislative action. I thank you all for your testimony today.
And I thank the audience for participating as well. Thank you,
Mr. Chair.
Chair Coons. Thank you, Senator Tillis. I understand
Senator Blumenthal would like a second-round question. Then I
have two or three, and then we will conclude. Senator
Blumenthal.
Senator Blumenthal. Yes. Thanks, Mr. Chairman. And I will
be very brief. I would ask each of you, not here, but in
writing afterward, to address two questions. One may seem
pretty simple. Mr. Salsberg, in your prepared remarks, you talk
about generative chemistry and how your company has developed
new potential treatments for malaria, which is extremely
promising. I would like from each of you--pick three, five of
the most positive developments, inventions, but emphasizing the
positive.
Because around here, we tend to be talking a lot about the
dangers and the harms because that's what attracts people's
attention, the potential for extinction. Not my word, but as
you know, a word used by many experts in the industry to raise
the potential for harm.
So, the second question I have for you, which doesn't
really relate to your reason for being here today, is jobs,
employment, impacts on our economy.
For me, what keeps me up at night, one of the main reasons
to stay awake when thinking about AI is the impact on our
economy. I think we're on the verge of another industrial
revolution, or the equivalent of the industrial revolution,
when many, many jobs were destroyed in the course of raising
the level of productivity for the jobs that were then created
as a result of industrialization. And the people who tried to
stop it, obviously the Luddites and so forth, failed.
But there was a lot of human impact there. Maybe if you
could address that issue and then maybe now, or in written
testimony--in written submissions. On the issue of open source,
Senator Hawley and I wrote to Meta earlier this week about an
AI model they released to the public in February. I'm sure
you're familiar with it, the LLaMA AI model. And open source
software obviously can be an enormously important resource for
innovation and invention, for science, standards, transparency.
It can foster competition. At the same time, even in the
near term, that kind of generative AI tool can be dangerously
abused. And we raised this issue in our letter to Meta. Once
the AI model and software are out there, there's no reeling it
back. And Meta didn't seem to put a lot of thought into the
risk before they released LLaMA. That's an issue, in my view.
Senator Hawley and my Subcommittee will consider this an issue,
open source, enormously promising, also potential dangers. I
don't know whether any of you now, or in writing, would like to
offer some observations.
Ms. Elluru. Thank you, Senator. We are working on assessing
the implications of open source models and trying to come up
with some policy recommendations and would be happy to get back
to your office on that.
Senator Blumenthal. Thank you. Any others? Ms. Sheridan,
did you have any thoughts?
Ms. Sheridan. Similar to what Ms. Elluru shared, everyone's
taking a look at this question right now and how it fits into
responsible AI innovation. So we'll certainly follow up with
more after the hearing.
Senator Blumenthal. But I think all of you would agree that
there's a need for thought and consideration here before these
models are just thrown into the wild, so to speak. Correct?
Thank you all. For the record----
Chair Coons. Thank you. Thank you, Senator Blumenthal. I'm
just going to ask three quick questions, then we'll conclude.
Ms. Elluru, if I might, just on the question of patent
eligibility, about Section 101 law, something we're going to be
turning to in a future hearing, how does the patent eligibility
status of core AI technologies and technologies applying AI
under U.S. law compare to eligibility under competitors Europe,
China? And is there a reason that the United States should
adopt a different eligibility test?
Is it necessary to harmonize eligibility laws globally as
we consider the role AI and other emergency--other emerging
technologies are playing in that race for critical technologies
in which AI is a critical, facilitating technology that you
described in your opening statement?
Ms. Elluru. That's a great question, Senator. Thank you. I
think, generally speaking, eligibility is not a hurdle in the
EU. Their law for exceptions is written pretty narrowly, and
it's interpreted narrowly. There's no dogma from courts
deciding what is patentable and what is not patentable. I think
China has a similar picture. They stopped--however, stopped
filing--publishing patent filing data in 2020. So it's hard to
get a picture of what's going on there.
However, I would say based on data from the last few years,
they were trending toward being more liberal in their
allowances for patent eligibility. That is not the case in the
United States. Since 2010, the patent eligibility uncertainty
has only expanded, and that's leading to unpredictability. And
so the patent incentives are----
Chair Coons. Eligibility has only expanded?
Ms. Elluru. Exceptions have only expanded.
Chair Coons. Exceptions have only expanded. Thank you. Just
wanted to make sure I understood.
Ms. Elluru. And so that is leading to uncertainty and
unpredictability and weakening the patent incentive. There is
no incentive to invent in the first place. And if there are
inventions, it's not shared with the public and is kept as a
trade secret. Either scenario is bad for innovation. As far as
harmonization, I think we need to be consistent with other
jurisdictions. I think different eligibility barriers for
companies is burdensome, which is not good for the innovation
ecosystem.
And the United States should not be an unwelcoming
jurisdiction with respect to patent eligibility. Patent
eligibility has implications on funding, job growth, the
economy. And so I think we need to be consistent, especially
with the partners and allies that we want to collaborate with.
Chair Coons. Thank you. Well stated. Briefly, if I could,
Professor Villasenor, Professor Abbott, is it something we
should be concerned about, that either a private actor or a
state actor is going to use AI to either write and file a very
large number of patent applications in an attempt to lock up
patenting opportunities, or will draft and publish a large
number of prior art publications to prevent inventors from
obtaining patents in the entire field?
Is this likely? Is it something we can forestall? And what,
if any, advice would you have for us on policy in this area, if
you would, Professor?
Professor Villasenor. So I do think it's a concern. I think
when there are the incentives in place which with the continued
advance of generative AI will be there, I think it's possible
that either a private entity and/or a state actor might do
exactly what you've asked in your question. The particular
policy responses depend specifically on what the pattern is.
It's very different if someone is publishing massive databases
of online preemptive prior art as opposed to filing massive
numbers of patent applications.
For the patent applications, there's some friction. It
costs money, for example, and patent--utility applications are
examined. But I do think it's sort of a watch this space area
where I do think we need to be agile with looking at how the
patent system might be used or misused in that way and
responding to it.
Chair Coons. And Ms. Sheridan, would the idea of raising
patent fees have any deterrent impact on this?
Ms. Sheridan. It certainly would. And with the caveat, of
course, that we're just talking about for large entities. But
yes, raising fees would disincentivize that sort of behavior.
Chair Coons. Professor Abbott.
Professor Abbott. Some slightly different thoughts on that.
I mean, if the concern is that AI is going to be generating so
much invention that it's filing a lot of applications on
legitimate ingenuity, I don't think that's a bad outcome. It
means we have too much good stuff. If an actor was deliberately
filing bogus patent applications to try and jam up the USPTO, I
think that's something we would have more of a technical
solution to. As to someone publishing a massive database----
Chair Coons. What would that technical solution be, sir?
Professor Abbott. Oh, well, I mean, if it was getting an
overwhelming number of kind of auto-generated applications
then, you know, the Patent Office is also leveraging its use of
AI and might have AI trying to detect kind of the use of
inappropriate filings that don't have legitimate claim and
specification structures in it, for example. You know, so
detecting essentially just fake applications. But you know,
that might be a complex subject.
To the extent, someone's just publishing a whole lot of
data to try and keep people from getting patents, that is
something I think patent law might have a solution to because
prior art has to be anticipatory and enabling. And I would say
if somewhere in a trillion, you know, document database,
somewhere there's some disclosure related to an application
that no one could usefully access or use, that would not be an
enabling anticipatory disclosure. But that might also be
something that if you are considering amending the Patent Act,
that Congress could clarify.
Chair Coons. Last question, Ms. Elluru, I think you raised,
but I've also heard about data being just as important when it
comes to AI as patents. And China allegedly is experimenting
with a new form of IP rights in data because of its relevance
to this race to be preeminent in AI. Should we be exploring
similar rights around data sets that would be designed to
incentivize innovation and promote disclosure and transparency,
some of the core goals of the patent law?
Ms. Elluru. That's a great question, Senator. Thank you
very much. In short, yes, data is a critical input to the
innovation ecosystem. We should be treating data across our
industry, public sector, and academia as a strategic asset. So
we need comprehensive data strategies to maximize the value of
that data. That comprehensive data strategy should include
several components, including data privacy, of course, but it
should also explore intellectual property rights and data.
Is it going to incentivize data-related inventions? Will it
incentivize data-driven products and services that are great
for the public? Is it going to help raise our economy--increase
our economy? It's estimated that the market for data in a few
years will be over a trillion dollars. Will intellectual
property rights and data help increase that economy? And I
think that is something we should explore.
Chair Coons. We might be able to continue that discussion
in our next hearing, which will be about the intersection of AI
and copyright. I am so grateful to Senator Tillis, his staff,
for making this possible. To all of you, for bearing with us,
for your great testimony. To the Members of the Committee who
came and asked questions, I will mention it's important,
technically, that I say Members can submit questions for the
record for you, the witnesses. They're due by 5 p.m. a week
from today on June 14th. And with that, this hearing is
adjourned.
[Whereupon, at 4:35 p.m., the hearing was adjourned.]
[Additional material submitted for the record follows.]
A P P E N D I X
Additional Material Submitted for the Record
[GRAPHIC(S) NOT AVAILABLE IN TIFF FORMAT]
[all]