Notes

Outline of the Class

• Recap on basic concepts of patent law
• AI-generated prior art (Can AI "destroy" the patent system?)
• Implications of AI for patentability requirements (Can AI qualify as a "tool"?)
• Implications of AI for patent inventorship and ownership (Can AI invent"?)

Recap on Patent Law - Patent Law Basics Patent

• Patent, title issued by an administrative authority, conferring the exclusive right of commercial exploitation of an invention for 20-years starting from the filing of the application.
• The rights are conferred only with the grant (following substantial/formal examination)
• Claims? Description?

Invention - 45 IPC / 52 EPC

• No general definition of invention in European Patent Convention (EPC) or Italian Industrial Property Code (IPC)

  European patents shall be granted for any inventions, in all fields of technology, provided that they are new, involve an inventive step and are susceptible of industrial application.

• Scholars: new and original solution to a technical problem, that is suitable for industrial application (functional definition).

• As long as it meets the mandatory requirements + subject-matter not excluded from patentability = patentable invention.

• Dynamic concept, to adapt to technological development.

• In the patentability requirements, the notion of "inventor" does not come up!

Novelty - 46-47 IPC / 54 EPC

• An invention shall be considered to be new if it does not form part of the state of the art (check the prior art!).
• The state of the art shall be held to comprise everything made available to the public by means of a written or oral description, by use, or in any other way, before the date of filing of the application.

Prior Art

• Priori art is any evidence that your invention is already known.
• It is enough that someone, somewhere, sometime previously has described or shown or made something that contains a use of technology that is very similar to the invention.
• A prehistoric cave painting can be prior art. A piece of technology that is centuries old can be prior art. A previously described idea that cannot possibly work can be prior art. Anything can be prior art.

Inventive Step - 48 IPC / 56 EPC An invention shall be considered as involving an inventive step if, having regard to the state of the art, it is not obvious to a person skilled in the art.

Person skilled in the art:

• Abstract mode that is only aware of the state of the art in the technical field of the invention (not "medicine", but "oncology", or - if applicable - "pediatric oncology) + common general knowledge (textbooks).
• May be a team, and thus be aware of multiple technical fields ("chemistry" + "oncology")
• May know other fields, if neighboring or even remote (respectively, if common to know or if prompted to do so).

Beware of the distinction! AI - Inventor vs AI - Invention

• Fundamental distinction between:
  • Invention/patent developed by an AI system (AI is the Inventor).
  • Invention/patent having an AI system as its subject matter (AI is the Invention).
• EPO new examination guidelines (into force on 1 November 2018). Essentially confirm the framework for computer-implemented inventions (CII).
• AI is a mathematical methods (MM). MM per se are devoid of "technical character" and thus not patentable. AI constitutes patenable subject-matter if it provides a technical solution to a technical problem:
  • An AI classifying text documents in respect of their textual. It is not an invention.
  • The use of a neural network in a heart-monitoring apparatus for the purpose of identifying irregular heartbeats makes a technical contribution. It is an invention.
• However, an AI system may invent a patentable AI system.

And now... to the real stuff Three hypothesis

Can AI "destroy" the patent system? AI-generated prior art

Can these texts constitute "prior art"?

• Remember: "Anything can be prior art". Potentially also automatically generated texts.
• Online content accessible through a specific URL is made publicly available if:
  • Can be found through a web search engine using keywords;
  • Accessible for a period of time adequate for direct and unambiguous access of the public.
• Cloem and the "All" project are both:
  • Published online in a non-confidential manner (potentially: "made available to the public") or filed at patent offices.
  • Time-stamped (established priority date).

Enablement, Technical Fields, Policy Risks

• On the side of the patent filing: Enabling Requirement: subject-matter described in a filing should provide information given to the skilled person which is sufficient to enable him to reproduce it.
• On the side of prior art: Machine-generated documents combining documents from remote technical fields could be disregarded (the reference could be considered not pertinent or analogous).
• Policy risks: monopoly in the hands of the automated-drafters, overburdening patent offices' workload, uncertainty for patent system.

For the time being very low likelihood of "destruction".

Can AI qualify as a "tool"? - AI as a tool in the invention process

• Super-tool?

  Dominant approach: AI cannot "invent" but is - at best - a sophisticated tool used by the inventors to reach the invention. Humans, not the machine, define the problem to be solved, set the parameters to solve it, identify the best solution (e.g. Oral-B CrossAction: selection of designs provided). "Human "ingenuity" is still necessary to define targets, parameters and success criteria" (Fraser, 2016). Different from Copyright? AI does not create the "full" output (but really?)

• Patentability

  The fact that an invention was made with the id of AI does not exclude its patentability. The way an invention has been realized is irrelevant to patentability. It is the result that counts, provided that it meets the patentability requirements.

• Novelty, there may be texts that are understandable/accessible by the AI and not by humans (e.g. coded, encrypted, mutated information). This could extend the universe of prior art to be considered when assessing novelty.

• Inventive Step

  • Relevant Prior Art, the skilled person as we know it:

    Will not consult prior art in remote technical fields (except if prompted to do so). Will consult prior art in a neighboring/broader general field only if the same/similar problem occurs (incentive to search in that field). AI is not biased to other technological fields, and may be able to factor in an enormous volume of technical data from remote fields. Need to broaden the scope of prior art? The broader the prior art, the more likely it is to find the invention lacking incentive step.

• Person Skilled in the Art

  AI may also affect the level of skill of the person skilled in the art. EPO: "the skilled person has at his disposal the means and the capacity for routine work and experimentation", which are normal in the field of technology. Is the use of AI a "routine" tool in the relevant field?

• NO, patent inventive If not obvious to the skilled person without use of AI (even if AI has done most of the work).

• YES, the skilled person is improved" with the AI. No inventive step if obvious to skilled person + AI (even if you didn't use AI).

  Yet, very complex to assess the capabilities of "normal" AI tool. How can patent examiners, patent attorneys and Judges assess this?

• Not disclosed in the application.

• Output difficult to predict (and may be different if the same input is fed to the system twice).

• Training data (standard/ not standard).

• Examined datasets.

• Self-learning and improving mechanisms.

  If the invention results from a large number of calculations, or brute force trial-and-error testing may seem non-obvious, but the result is inevitable?

• Industrial Application - Person skilled in the art

Can AI "invent"? - AI as an inventor

• Does AI invent?

  • Position 1: Some argue that autonomous AI inventions are still far away, but address the possibility (cf. antenna in deep space example).
  • Position 2: Others are convinced that, due to the minimal or exceedingly fragmented role of natural persons, inventions are already being developed (essentially/primarily) by AI systems autonomously.
  • Position 3: More reasonably, even if we were to accept that AI invents, the AI is a "joint" inventor, within a skilled research team (still designed, developed, prompted and assessed by humans, but play along like a member of the team), does not invent in a vacuum (yet).

• Patentability: is an AI-invention actually an invention?

  Not anthropocentric view of invention (although hints of "inventor's creativity" in case law and scholarly approaches in EPO Member States).

• Different from US: patentable subject-matter "anything under the sun made by man" (Diamond v Chakrabarty).

• But: recent position of the UKIPO would not allow to list an AI Inventor.

• Different from Copyright: "originality" is defined as "author's own intellectual creation" (IT, also: "expression of the author's personality..."); "author" is the foundation of copyright law.

• Preliminary Distinction

  • Inventorship vs. Ownership

    The following applies both for AI as a "tool" and AI as an "inventor".

• Inventor

  No definition of inventor in the EPC or IPC. However, the inventor is generally framed as a human being.

• Art. 60(2) EPC: "If two or more persons have made an invention independently of each other..."

• Art. 185 IPC: the patent application must include first and last name of the inventor.

• Art. 62 IPC: the moral right to inventorship may be enforced by the inventor itself and - after its death - by its spouse, its descendants, its ascendants or, in lack thereof, by other relatives up to the fourth degree.

  So? It is not possible to claim an AI as the inventor in the current framework.

• What happens if the applicant omits the inventor?

  Art. 81(1) EPC: "The European patent application shall designate the inventor". If the designation of the inventor is absent or deficient and the inventor does not provide the name within 16 months from the filing/priority date the application is refused (Art. 60 + 90(5) EPC). However, if the missing/incorrect designation goes unnoticed: this is not a cause of revocation of the patent (numerus clausus rule) (Banterle 2018).

• Potential solutions to inventorship issues?

  • No patent protection for AI inventions.
  • AI as "tool" - need to find a "peripheral" human inventor.
  • Amendment to encompass AI as the inventor.

• No Protection

  • Ravid-Liu (2018), When Artificial Intelligence Systems Produce Inventions: An Alternative Model For Patent Law At The 3A Era:

    "We argue that [...] these inventions should not be patentable at all, and that other tools can achieve the same eds while promoting innovation and public disclosure. These other, non-patent incentives include commercial tools such as electronic and cyber controls over inventions, first-mover market advantages, and license agreements. [...] In maintaining the traditional patents system by hunting for a "real" human inventor, policy makers exhibit a misunderstanding of advanced technology and AI system features".

• Peripheral natural inventors "behind" the AI

  • The AI user.
  • The AI programmer (trainer, provider of data, software developer, ...).
  • The AI owner.
  • The person that recognizes that the AI's output is a patentable invention.
  • And potentially others... possibly too fragmented? AI is now being developed by very large research groups.
  • Who do you think is best placed to be named as the inventor?

• AI recognized as the inventor

  The Ai (in this scenario) contributes and communicates all the elements of the inventive concept, it can be seen as the inventor. Most likely, Ai inventions are developed within a team anyhow, so the AI would be one of the members of the team. However, this requires a human-neutral interpretation of patent concepts.

• Ownership

  Following the discussion on inventorship -> ownership of the patent. The general rule is that the inventor is the initial owner of the right to paten, the right "originates" upon him (Art. 63 IPC - 60 EPC).

• Ownership - Employed-to-Invent

  • Art. 64 IPC: Inventions made by employees (EPC leaves this matter to Member States).
  • The right to patent belongs to the employer when the invention is developed in performance of an employment contract.
  • Rationale: the relevance of economic investments in R&D is bigger than that of the inventor's role in developing the invention.
  • The same applies to commissioned inventions (no employment relationship).

• Employed-to-Invent and Commissioned Inventions

  • Position 1: the right to patent inventions developed within employed (or commissioned) work "originates" directly upon the employer/contracting party.
  • Position 2: the right to patent developed within employed (or commissioned) work "originates" upon the inventor and then is automatically transferred to the employer/contracting party.

• AI as Employed-to-Invent

  • The employed-to-invent principles could be adapted to the scenario where AI is indicated as the/amongst inventor(s).
  • If one were to consider the AI as an employee/contractor there would be no issue of assignment of rights, of any kind, as the rights to patent would "originate" directly upon the owner/customer, while the recognition of moral rights is inherently excluded.
  • This requires two interpretative" hops (AI as inventor and AI as employee) but would seem to generate the maximum benefits in policy objectives and legal certainty for third parties, so we shall wait to see what happens next.

• Policy Considerations

  • Cons of not allowing AI inventions
    • The AI's role in the invention may be omitted in the patent application (see Thaler's patent); undermines certainty in the system and causes a fracture from reality in the assessment of the patentability requirements (e.g. factor the use of AI, or not, in a technical field).
    • May discourage innovators from pursuing AI inventions (which still require humans to create and maintain AI systems), thus stifling tech-progress.
    • AI-inventions may be kept as trade secrets (non-disclosure instead of disclosure; risk of a perpetual monopoly): again, risk of stifling innovation.
  • Pros of not allowing AI inventions

    Would prevent a number of risks:

• Impact on future human innovation since supplanting human intelligence may result in atrophy of human intelligence and obsolescence of the scientific method;

• R&D job market (innovation is more accessible);

• Risk of monopolies in small fields (with innovation at very quick rates).

• Dominant Approach

  IP protection for AI-Innovations is justified (utilitarian approach: as long as the inventions meet the patentability requirements, i.e. the contribute to progress).