The case for
reforming the concept of Inventorship under the UK Patents Act 1977.
| (Photo: PhonlamaiPhoto / iStock / Getty Images Plus) |
In December 2021, the UK Government published its national AI strategy which
highlighted its goal to become the leading nation in the development of
Artificial intelligence. As part of the AI strategy, the Government recognises
the importance of a progressive regulatory environment which encourages
innovation, promotes the use of AI for the public good, preserves the integral role
of intellectual property – the need to promote and maintain human creativity
and it must do so based on the best available economic evidence.
The UK Government Consultation on reforms to copyright and patent law to
incentivise the development of AI closed earlier this past month on 7 January
2022.
The issue of Patent law and AI, namely AI inventorship has raised an
ongoing global debate in the patent world, through the case of
Thaler v The Comptroller, also known as DABUS. This article sets out
the current law, provides a critique of the current law, proposes the need to
reconsider the concept of inventorship and provides a rationale evaluation of
the proposals put forward.
On 21 September 2021, the Court of Appeal of England and Wales delivered
the awaited judgement of Thaler v The
Comptroller [2021]. By a majority 2–1, (Arnold and Laing LJJ in the
majority and Birss LJ dissenting), it was decided Thaler had failed to process
his patent applications correctly as required by s13(2) of the Patents Act 1977, (the ‘Act’).
Thaler had identified DABUS: [The Device for the Autonomous Bootstrapping of
Unified Sentience], (the AI creativity machine created by him) as the sole
inventor of two inventions.
Arnold LJ stated it was a legal impossibility [para 143], the rights of an
inventor do not extend to an AI under the Act. AI creativity systems such as
DABUS do not have legal personality or capacity and thus do not have a property
right in inventions created by it nor can DABUS enter into a transaction with
Thaler to transfer title to him.
The rationale of the judgement based on the Current Law
Defining the Inventor
s7(3) defines the inventor as the “actual deviser” of the invention. The
natural person who “came up with the inventive concept”, [University of Southampton’s
Applications [2005] RPC 220 [234] Justice Laddie]. Lord Hoffman
in Yeda [para 18] stated that the 1977 Act is a
‘complete code’ because entitlement to a patent is dependent on being the
inventor. A patent may be granted “primarily” to the inventor under s7(2)(a) or
someone claiming through him via Ss. 7(2)(b) and (c) which mention the wording
“person or persons”, such as an employer, s.39 PA 1977. s13(2)(a) requires the
applicant to “identify the person or persons whom he believes to be the
inventor.”
The operation of s13
The purpose of s13(2) is to ensure the inventor gives the Comptroller
information required to be made public on the register. The role of the
Comptroller is “to check whether the prescribed statements have been filed in
time, and if so whether the statements filed appear to comply with the
statutory requirements or are defective on their face”,[Nippon Piston Ring
Co.’s Application [1987] RPC 120]. The current position in the law is that
there is a higher obligation of genuine belief imposed on the applicant to
accurately name a “person” as an inventor and to indicate derivation of rights
to avoid non-compliance as a result of defective statements under s13(2).
Critique of the Current Law + Identifying the Problem
The effectiveness of the law rests on its
ability to balance the interests of the inventor on the one hand, and the
interests of the public on the other.
The ‘person’
International Harmonisation
Prohibiting inventorship of non–persons under s7(3) maintains
consistency in the law. A recent 2019 study on
inventorship and AI conducted by the EPO found that the
inventor in most jurisdictions must be a natural person. This creates certainty
for potential applicants as seen through the stream of refusals obtained by
Thaler across jurisdictions. According to CIPA’s Computer
Technology Committee Consultation outcome, many companies
wish legislators to ensure a cautious and predictable approach with significant
human involvement. The current law reflects societal concerns about the
importance of maintaining human involvement.
Obviousness
s7(3) provides stability because it prevents future straining of the
fundamental building blocks of the patent system — the development of an
inventive idea sufficiently enabled to use by others to solve a technical
problem, — the inventive concept — demonstrating technical creativity. This is
because the current use of ai technology systems is not completely autonomous
since data still has to be inputted, from which the Ai machine can perform the
inventive step. The narrow scope of s7(3) means that refusing legal personhood
to Ai systems avoids confusion with regard to the threshold of the inventive step.
If DABUS were given legal personality
then inventions it would subsequently invent could still be seen as inventive,
because obviousness is judged through the eyes of the Person skilled in the
Art. As the skilled person is taken to have read all literature, possessing
common general knowledge, [Rockwater v Technip
France SA [2004] RPC 46 (CA) [Para 7] Lord Justice Jacob].
It is settled that
this man, if real, would be very boring — a nerd. Lord Reid put it this way in
Technograph v Mills & Rockley [1972] RPC 346 at p.355
“… the hypothetical addressee is a skilled technician who is well
acquainted with workshop technique and who has carefully read the relevant
literature. He is supposed to have an unlimited capacity to assimilate the
contents of, it may be, scores of specifications but to be incapable of a scintilla
of the invention. When dealing with obviousness, unlike novelty, it is permissible
to make a “mosaic” out of the relevant documents, but it must be a mosaic which
can be put together by an unimaginative man with no inventive capacity.”
The Person Skilled in the Art would already have knowledge of the
existing field of AI. s7(3) prevents future conceptual uncertainty on what is
routine information to an AI machine which would raise the threshold of
inventive step. Although, if as predicted by Abbott AI systems
reach the level of human intelligence then all inventions would be non-obvious,
and the threshold of inventive step would be too low.
The other side of the Coin
The narrow scope of s7(3) applied in Thaler has created the dilemma of
an invention without an inventor. On the one hand, the existing law makes it
difficult to patent an invention in which the human aspect is limited to “mere
application of time, money and effort” [Bently and Sherman,
page 960], with a limited technical contribution. This is the current law benefit
for society as the non-patentable ai generated outputs are in the public
domain. Balladrini justifies this outcome of the
current law by stating that ai does not need to be rewarded for its output
since it does not possess human incentives of money and recognition. However,
the human who has created the machine is encouraged to innovate in exchange for
a reward. Consequently, the inventor is disincentivised to disclose their
invention and becomes encouraged to gain alternative protection via a trade
secret, subsequently harming the public interest. This risks crippling
innovation and development. Courts have acknowledged that the technical
feasibility of an invention is unlikely to be isolated from what is commercially
feasible as illustrated in Dyson, although the
exclusion of non-human inventors makes the current law outdated when compared
to existing new technologies by not also equating the economic right to the
investment inputted by the human.
Transparency: Interpreting s13(2)
The application and ambiguity in the interpretation of compliance with
s13(2) in its current state threaten the transparency of the patent system. On
the one hand, Arnold, and Laing LJJ in Thaler uphold the strict literal
approach by placing a higher threshold of “genuine” belief to identify the
inventor requirement as an extension of s13(1) (attribution right), which gives
effect to the paternity right of the inventor, rights which are unattainable by
DABUS. The presumption of s7(4) inventorship is agreed upon. In contrast, Birss
LJ concludes from Nippon that Thaler merely stating his right under s13(2)(b)
“by assignment” is sufficient in the eyes of the office and there is no need to
indicate which of ss.7(2)(b) or (c) apply. The differing interpretation of the
s13(2) formality suggested by Birss LJ is that the role of the comptroller is
not to adjudicate on the effectiveness of the claimed derivation of right.
The consequence
Both interpretations demonstrate that the entitlement of the applicant
rests on whether the inventor is a person and ownership is distinct from
inventorship. There is no law in place to transfer the right from an AI since
it was not in the minds of parliament when reforming the 1949 act, since AI was in its
infancy. The existing hurdle of proving the s7(4) presumption to prevent a
later revocation under s72(1)(b) allows human creators of ai creativity
machines to circumvent the human inventor requirement by taking credit for the
inventions of the ai system. It creates “a proxy human
inventor” who is not the actual deviser. For example, John Koza earned a US patent for a
system designed to make factories more efficient, generated “without human
intervention and in a single pass” by his AI “invention machine.”
If he had told the truth, the invention would not have been patentable on
grounds of lack of human inventorship. Thaler also has existing
patents for; the cross-bristle design of the Oral — B Cross Action toothbrush,
new super-strong materials and devices that search the internet for messages
from terrorists, [Page 1087]. These patents were granted by the
USPTO, however, Birss LJ in the case of Thaler v The Comptroller—General of
Patents, Designs and Trade Marks [2021] EWCA Civ 1374 [para 81] reaffirmed that
had Thaler named himself as the inventor, there would be no problem.
 |
©️ Ryan Abbot: This diagram shows how a container's shape could be based
on fractals - Application for FOOD
CONTAINER Patent
Section 13(2) considers if there has been a bona fide mistake on the
part of the applicant in filing applications. It no longer allows for the “true
inventor” to be identified because the law does not have a mechanism for
identifying a non–human inventor. This hinders recognition of the contribution
provided by the ai creativity machine. The ultimate objective of the patent act
to reward disclosure of a technical contribution is not met or achievable
rather the integrity of the system is distorted with the dishonest inventor
reaping from what he has not sown for a term of 20 years.
Accountability through Ownership
The current states of ss.7(2)(b) and (c) have been interpreted by courts
to establish clarity on ownership of inventions where AI has been used as a
tool in the invention process, [HTC Europe Co. Ltd. v
Apple Inc. [2013] EWCA Civ 451]. Although, inflexible when applied
to AI-generated inventions the current law holds inventors and persons to whom
the right has been assigned via s7(2)(b) and (c) accountable in the case of
potential infringement. Ownership for AI system inventions is limited in scope
because of s7(3) which leads to the same consequences discussed above.
The Need for reform — Summarising the Problem with the Current Law
The current scope of inventorship under Sections 7 and 13 is too narrow.
The existing law demonstrates a necessity for a human rights holder but
through a mechanism not provided by the Patents Act 1977. The reform proposals
below aim to meet the goal of providing stability and ensuring the upholding of
the Patents system values of integrity, accountability, and transparency. The
notion of the inventive concept should be protected by considering the
investment made and the nature of the industry to determine the scope and term
of protection.
Proposing a Sui generis system 💡
I propose a sui generis right for AI-generated inventions to run
alongside the Patents Act 1977 consisting of four components.
I. The owner of the Ai machine (the person who created the ai generated
output) should be derived ownership, in a manner similar to the sui generis
protection afforded under Recital 41 of the
Software database directive.
EU Directive
96/9/EC on the Legal Protection of Databases:
Recital 41: (41)
Whereas the objective of the sui generis right is to give the maker of a
database the option of preventing the unauthorized extraction and/or
re-utilization of all or a substantial part of the contents of that database;
whereas the maker of a database is the person who takes the initiative and the
risk of investing; whereas this excludes subcontractors in particular from the
definition of the maker;
II. The patentability criteria in Section 1 of the Patents Act 1977 are
to apply. The useful technical effect” test for the inventive step is to be
considered through the lens of the commercial viability of the invention taking
into account the market and the right holder’s disclosure of the data inputted
to demonstrate their substantial contribution, which is to be assessed by the
person skilled in the art, (experts in the field of AI and the industry which
the invention concerns).
III. The term of protection for AI-generated inventions is to be reduced
to 10 years term of protection.
IV. The rights conferred are to be economic, not moral rights.
Rationale Evaluation of the Reform Proposals [I-IV]
I. Ownership
The sui generis approach removes the need to find an inventor, adopting
the database system method in which the maker of the database (the creator of
the ai machine) and the person who invests money and time in it is to be the
right holder. This approach provides clarity on entitlement to the AI-generated
invention. The main disadvantage is that it does not guarantee that a human
inventor will not take credit for an AI-generated invention via the traditional
patents system route, to benefit from 20 years of term protection, in
comparison to 10.
Davies proposes the
use of contractual protections for ai generated outputs through private
contracts allowing private investors to decide on how they wish to apply AI
outputs. This proposal is not practically workable, economic rights attached to
the AI invention outputs would be resting on contractual freedoms. This
promotes inaccessibility for SMEs and imbalances in potential bargaining power.
It means there will not be any consistency in the law and the disclosure of
technical progress to society risks being hindered by over competitiveness
leading to secrecy. In comparison, the sui generis regime is accessible and
encourages investment.
The Australian AI Inventor
Section 7 of the UK Patents Act 1977 provides a narrow scope of who is
an inventor, the definition which if expanded can extend the scope of
ownership. The implications of giving legal personality to an AI creativity machine
are illustrated in the Australian Case of Thaler v Commissioner
of Patents [2021] FCA 879. Beach J considers “inventor” as an
agent noun which he compared with the widened concept of “manner of
manufacture” in light of new technologies, terms which originate from Section 6 of the
Statute of Monopolies 1623. The ramifications of the legal
personality approach may include the Person skilled in the Art being considered
as a person assisted by Ai, and the Common General Knowledge expanding to
include AI developments, complex ownership issues may arise, as well as
uncertainties around disclosure on the specification leading to an influx of
patent applications from many industries. Modern AI machines have not advanced
significantly from the 1970s expert system because both systems function
through inputting of data, the distinguishing factor is the potential
unpredictability of the modern ai machines. The proposed sui generis regime
provides a medium between the human-only current UK approach and the
shoehorning into existing legislation and benefits theoretically from not
dealing with the aforementioned issues.
II. Inventive Step ➡️ “Substantial
Contribution”
This sui generis regime adopts the term “substantial contribution” as a
mechanism to measure the impact of the human who utilises an ai machine by
inputting data. It is already recognised that commercial and technical
considerations can be intrinsically linked, particularly in the case of Ai
inventions. The proposed definition of substantial contribution should depend
on how the right holder gained access to the data inputted and if it was
collected and obtained through skill and labour with a significant amount of
time and effort. The benefit of defining human contribution maintains the
strength of inventive step, unlike the disappearing requirement of originality
for AI-generated copyright works under s9(3) CDPA 1988. There is limited
jurisprudence or discussion on this matter with regard to patents. However,
the analogous copyright demonstrates the effects of only creating a test —
“taking arrangements necessary” — without any form of enforcement. The
application of s9(3) CDPA, as put forth by Ginsburg, is “an
unfortunate and confusing conflation of the concepts of authorship and
ownership.” The same issue is present in the current UK Patents Act 1977.
However, the sui generis regime creates a requirement of disclosure to evidence
the contribution. The benefit is that only the right holders who sufficiently
disclose their contribution can obtain protection.
Disclosure
Disclosure is the patent bargain through which inventors reveal
sufficient information to allow the Person Skilled in the Art to create the
invention without the need to adopt inventiveness. The human rights holder
would be required to describe the steps taken in the process by the AI machine
and the significance of their role. The rationale behind this component is to
maintain the human element and to bring transparency to the patent system
through clarity of the inventive concept, by assessing whether the activities
which the human has disclosed as being conducted by him are actually of
creative and intellectually inventive quality in relation to the nature of the
technical features of the subject matter. This means if an invention is found
to be non-obvious, then the disclosure will determine if it was due to the
steps taken by the AI machine. The problem in implementing this would be the
inevitable issue of the black box. As the input and output for an ai machine are
clear but the difficulty of establishing the in-between may be difficult to
explain. There are practical limitations to achieving transparency in the
proposed regime and existing law.
Stanková proposes
enhancing transparency by involving scrutiny from competitors and the public to
challenge the accuracy of disclosure on grounds of unpatentability of subject
matter. Although this approach enforces accountability, it does not seem
commercially sound to have your competitors challenge the validity of your
invention. Yet it may prevent a surge of applications and achieve its aim of
transparency allowing protection via merit.
III. Term of Protection
The 10-year term makes protection under this sui generis regime
appealing to the owner of the creativity machine who cannot patent their
invention under the existing Patent law or Trade secrets. Trade secret
protection ends once the invention loses its commercial value and this could be
earlier than 10 years, especially if the novelty of the invention is visible
because it could be reverse-engineered. The additional obligation of preserving
secrecy means that protection via the trade secrets route may not fully
appreciate the value of the AI-generated invention. It is estimated the AI
market will add more than $15 trillion to the global economy by 2030. The reduced term
to 10 years is because the human does not complete the inventive step by
providing a technical contribution. This mechanism would be similar to the
shortened protection term of 50 years for ai generated copyright works in which
the deemed author takes the “necessary arrangements”. The proposed reform and
s9(3) CDPA 1988 are able to promote societal welfare without costing the right
holder of his reward.
Lauber — Rönsberg and
Sven put forth that the data obtained and inputted into ai creativity
machines should be considered “a collection of data” which should be protected
by the database directive, for a duration of 15 years. Lauber — Rönsberg and
Sven are correct to identify ai generated inventions as “information” but as
Lord Hoffman established in Merrell Dow v Norton
[1996] RPC 76 inventions that are not merely information can be a product, such
as the beverage container and neural flame devise created by DABUS. The
database approach would be inconsistent in protecting all AI-generated
inventions, especially because the data is processed by the AI machine to
invent, and such information is not software. The 15 years duration is too long
because AI technology develops rapidly, and the benefit gained from a 15-year
limited exclusive right would not equate to the technical consideration aspect
used to measure “substantial contribution” and would have a chilling effect on
innovation.
The proposed 10-year term of protection begins from the point of filing
and will be six years once the patent has been granted taking into account the
nature of the invention if it is a product and a reasonable time to recoup the
investment. The measure of reasonable can be evidenced by the average holding
period of a private equity fund being 5.4 years. Yet there is
no strict method available on how to measure the optimum duration.
IV. Removal of Moral rights
The sui generis right afforded to owners of the AI creativity machine
would not require an assignment of any rights because a machine is not morally
incentivised to seek attribution. The moral rights of the human inventor to be
named are engrained in s13(2). The goal of a sui generis system is to encourage
investment and prohibit market failure. Stephens argues
‘Changing patents into a pure economic right would have a significant impact on
the system with questions on whether society wants to reward owners and
investors in ai systems.’ There is merit to Stephens’ argument because the
complete removal of all moral rights could also lead to a lack of accessibility
for smaller and medium enterprises that may not have an investor base.
The solution is in the interpretation of the substantial contribution
(component II) provided and a significantly shorter term of protection for
rights that cannot be licenced. Taking into account the limitations,
competition law principles enforced by the CMA would apply in order to provide
added protection to the consumer, ensuring a balance between societal benefit
and the right holder’s interest. Patents are industrial property, and unfair
competition principles under the Paris Convention, namely Article 10bis to
ensure consumers are effectively protected against any dishonest practices
would also apply.
Conclusion — A need to take action
The proposed reform puts forward that the sui generis right replaces the
human inventor with an accountable human rights holder who is the owner of the
AI creativity machine. The aim is to ensure a balance of the interests of the
public with the right holder whilst theoretically ensuring a fair and
transparent system. The true effectiveness of any law is only known upon
implementation. It remains to be seen what path the UK takes following the
closing of the UK Government Consultation last month on 7 January 2022,
as the actions taken are
fundamental for the future of its innovation and economic growth to become the
global leader in AI.
The UK is not alone
in declaring its intention to become “ a principal world centre of Artificial
intelligence innovation” as China has also vowed to do by 2030. The race to
dominate the AI industry now includes the UK, China, Germany, the United States
and Japan. Whether we can soon see a harmonised approach for AI protection
through international cooperation in the near future remains to be seen.
