11: AI & IP Flashcards
(16 cards)
What is intellectual property?
IP protects immaterial output (form, shape, lyrics, song, book)
=> Grants a negative right to the creator/author: a right to exclude others to doing things with the thing you created (negative, so you don’t have to do things with it)
What is trademark and design law?
Trademark laws: logo’s, brands, things used to distinguish goods or services between one actor and the other
Design law: protects distinctive shape or form of a good
Difference between AI-generated and AI-assisted output?
AI-generated: no to minimal human intervention, no relevant human contribution to the output
AI-assisted: more relevant human intervention in the creative or inventive process
What is copyright?
- original literacy or artistic works: About works not about ideas => has to be concretely expressed in a text, song, painting..
- authorship
- economic and moral rights:
-> Economic right: allow for copying or further dissemintation => can be the subject of a contract
-> Moral right: right to paternity = right to ask to mention your name with your work
-> Right of integrity: avoid that others hurt the work you created by making adaptations to your work - no formalities: copyright applies when there is work that is original, no registration or approval needed
Scope of copyright for AI-technology?
In scope:
* preparatory design materials: necessary materials leading to the creation of a computer program
Ex. flow charts, descriptions, program ideas..
* source code
* written form of program code
* object code
Out of scope:
* computer program functionality
* programming language
* mathematical formulas
* data formats
Can someone become the author of AI-generated works?
If we did not know that AI was involved we would grant copyright to it => now that we know that AI is involved, will our assessment be different?
Divide in authorship and ownership. This looks at authorship
=> Variety of considerations that prevent giving copyright to AI-generated works (no human involvement):
1) Originality requires free and personal choices of an author
=> author imprints his personality on the output
=> AI systems do not have a personality: AI can not fulfill the originality constraint
2) Copyright is a droit d’auteur: auteursrecht, author is the primary actor in the copyright
=> this does not allow for other entities (like AI) to play a role in that
3) Copyright is a human right
=> not able to grant it to an artificial entity in the current law landscape
4) Author as first owner: first owner of copyright protected work is the natural person that created the work (refers to the fact that it has to be a natural person, not an artificial entity or a legal person (company))
5) Term of protection: copyright lasts for 70 years after the dead of the author
=> cannot accomodate for this term of artificial entitites
6) Copyright contract law: protecting the author (human being)
7) Moral rights can’t be applied to artificial entities
=> author as established under existing copyright law should be a human and does not allow for other entities such as AI to be considered as an author
=> only protect works created with AI systems if there is sufficient human interaction => need for the human actor
==> copyright protection for AI-assisted works but not for AI-generated works
What are the considerations at stake when we think about granting AI-generated works copyright?
1) To who should we grant the copyright => AI system does not have a legal personality and wield rights
2) Hypothesis of AI-generated rather small, so a lot of works will fall under AI-assisted and will be grand copyright so why consider AI-generated work
3) No proof of incentives: grant copyright to the AI-system but not to the developer, so the developer is not incentivised to further develop AI-system because it does not enjoy copyright protection
4) Effect on freedom to create: if AI-generated work gets copyright then a lot of AI-generated works will be created, so the market for human creators will be less and less big => because AI-generated much easier and faster
5) Modalities: how will the copyright be applied in practice (also 70 years after the deactivation of the AI-system, what kind of rights (only economic rights?))
6) Ownership debate: to whom do we grant the rights if not to the AI-system? Developer, provider of training data, user of the system…
What are the legal based that can be used to train an AI-system under copyright?
3 legal bases that can be used to train an AI-system legitimately under copyright in Europe:
1) Ask author for his permission: can your works be used to train an AI-system?
=> Lot of work, solution: managed by collective management organisations
2) Text and data mining exception
3) Use public domain data (Ex. Rembrandt): Rembrandt passed a way long time ago so his works are in the public domain so it can be used freely
=> Risk: using outdated, biased data
What are the text and data mining exemptions? (TDM)
Copyright material cannot be used unless permission or exception
2 exceptions:
1) RO: research organisations and CHI: cultural heritage institutions
=> Limitation: what is created due to TDM cannot be commercialised
2) limited, revocable exception for everyone else if the rightsholder has not expressively reserved its right
=> You can use all the copyright material unless the right holder reserved its rights (indicated that you cannot use it)
=> a lot of material can be freely used
Patent law?
- inventions/technical advantages: Patents protect inventions
-> inventions must provide a technical advantage over the preexisting state-of-the art
=> Also a negative right - temporary and territorial: Patents last for 20 years and only applies in the territory of its jurisdiction
- economic and moral rights:
-> Economic right: similar to copyright law, ensuring that others don’t recreate what you invented
-> Moral rights: paternity and integrity - registration: different from copyright, you have to register your invention
=> disclosure: only allow patent protection if the technical knowhow is disclosed to the public to enable scientific and technical progress
Patentability requirements invention
1) no excluded subject matter
2) novel
3) inventive
4) industrially applicable
Patentability in AI-context
- lot of creations are excluded from patentability because they lack technical character
- mathematical methods (models and algorithms) and programs for computers (software) are excluded and fall under excluded subject matters => no patent
HOWEVER,
-> MM including technical device: computer, robot, actuator
=> no exclusion so could be patented
-> Computer Implemented Inventions: computer program should create a further technical effect
-> if you can prove this effect, the exclusion does not apply
What is a further technical effect?
FTE: TE going beyond the normal physical interactions between program and computer
=> Normal physical effect: Circulation of electrical currents within a device
=> FTE: when it controls technical processing or internal technical functioning of a device
* Control of a technical processing: in order to 3D print, you need a CAT file (visual representation of the object) => new CAT file that also included the technical instructions that the 3D printer will use immeadiately, this type of CAT file will control the technical process and can be considered FTE
* Internal technical functioning of a device: opening/closing a switch on a circuit board in a better/more efficient way
What is an alternative approach to get patented?
referring to ‘any technical means’:
Just by refering to it can be used to escape the excluded subject matter and could attribute technical character to the claim
However, only technical features contributing to the technical solution of the problem are relevant to assessing the inventive step required to get patented, so if you merely included technical means in your claim but it doesn’t attribute to the technical solution to the problem, then you will have a problem with proving the inventive step
What is patent bargain?
Patent protection in exchange for public disclosure of invention
Patentability of AI-generated output?
1) human inventorship applied in patent legislation
2) first owner is either the inventor or the employer
=> AI has no legal personality so cannot be the owner or the employer => AI cannot be the inventor
3) justification of IP protection are tailored to human authors/inventors
==> no AI-inventorship, AI-generated output is not patentable
AI-assisted output is patentable if there is sufficient human contribution
