Ar dirbtinis intelektas, naudodamas kitų menininkų kūrinius naujiems kurti, nepažeidžia autorių intelektinės nuosavybės teisių?
Rusteikaitė, Ada |
Sparčiai tobulėjant dirbtinio intelekto technologijoms, skaitmeninėje aplinkoje prieinamos programos geba sugeneruoti unikalius meno kūrinius – poezijos, dailės, fotografijos, muzikos srityse. Šių sistemų duomenų bazės sudaromos iš didelio masto kitų menininkų kūrinių, kuriems taikoma autorinių teisių apsauga, juos toliau naudojant mašininio mokymo procese. Programų vystytojai – paslaugų teikėjai, naudodami autorinį turinį skaitmeninėje aplinkoje yra įpareigoti užtikrinti naudojamų kūrinių teisinę apsaugą. Dauguma programų netaiko licencinių sutarčių sudarymo praktikos su autoriais, kurių kūriniai naudojami, dalis nenumato pareigos nei sau, nei vartotojams sumokėti jokio licencinio mokesčio kūrėjams, kurių kūriniai buvo panaudoti, įmonės pelnosi iš tokios veiklos gaudamos net milijardinius finansavimus. Neaiški atskirtis, ar dirbtinis intelektas gali būti laikomas autoriumi, o jo kūrinys – individualia kūrybine išraiška, kuriai taikoma apsauga. Visa tai kelia autorinių kūrinių naudojimo teisėtumo klausimus. Darbo tikslas - išanalizavus kūrinių naudojimo dirbtinio intelekto meno programose teisinį santykį, nacionalinius, tarptautinius teisės aktus bei teismų praktiką, atskleisti autorinių kūrinių naudojimo šiose programose probleminius aspektus ir įvertinti, ar toks autorinių kūrinių naudojimas nepažeidžia autorių intelektinės nuosavybės teisių. Darbe naudojami aprašomasis, lyginamasis, literatūros, teisinių dokumentų, teismų praktikos, loginės analizės, metodai. Tyrimui pasitelktos naujausios bylos ir Direktyva 2019/790, kuri viena pirmųjų pasaulyje reglamentuoja gaires bei apribojimus autorinių kūrinių naudojimui mašininio mokymo procesuose. Pastebėta, kad dirbtinis intelektas neatitinka autorių teisių subjekto sąvokos, kas sukuria dvi problemas: 1) subjektiškumo (programa gali būti kūrėju, tačiau ne autoriumi); 2) objektiškumo (kadangi programa negali būti autoriumi, toks kūrinys negali būti apsaugos objektu). Autorių teisių pažeidimų analizė atskleidė, kad pagal Direktyvą 2019/790 autorinių kūrinių naudojimas be sutikimo mašininio mokymo procese atitinka mokslinių tyrimų kategoriją ir yra teisėtas. Toks naudojimas taip pat atitinka sąžiningo naudojimo doktrinos principą tais atvejais, kai nesugeneruojama neteisėta kopija ar vertimas, kūriniai yra transformuojami į naują individualų produktą, naudojami mokymo tikslais, neturi neigiamos įtakos kūrinio vertei rinkoje, tačiau gali pažeisti turtines ir neturtines teises, kai sugeneruojama neteisėta kopija ar tiesioginis vertimas.
With the rapid development of artificial intelligence technologies, art programs available in the digital environment are able to generate unique works of art - in the fields of poetry, art, photography, music. The databases of these programs are built from large-scale collections of copyrighted works by other artists, which are further used in the machine learning process. Program developers - service providers, using copyrighted content in the digital environment, are obliged to ensure legal protection of the works used. Most of the programs do not apply the practice of concluding license agreements with the authors whose works are used, some do not require either themselves or the users to pay any license fee to the creators whose works are used, companies profit from such activities by receiving even billions of funding. The distinction between whether an artificial intelligence can be considered an author and its work an individual creative expression subject to protection is unclear. All this raise questions about the legality of using copyrighted works. The purpose of the work is to analyze the legal relationship of the use of works in artificial intelligence art programs, national and international legislation and court practice, to reveal the problematic aspects of the use of copyrighted works in these programs and to assess whether such use of copyrighted works does not violate the authors' intellectual property rights. Descriptive, comparative, literature, legal documents, court practice, logical analysis, methods are used in the work. Recent cases analyzing the objectivity, subjectivity and use of works without the authors' consent were used for the research. Directive 2019/790, which is one of the first in the world to regulate guidelines and restrictions for the use of copyrighted works in machine learning processes, were also extensively analysed. It was noted that artificial intelligence does not correspond to the concept of a copyright subject, which creates two problems: 1) subjectivity (a program can be a creator, but not an author); 2) objectivity (since the program cannot be the author, such a work cannot be the object of protection). A work of artificial intelligence is characterized by originality, but is not yet recognized as an object of copyright, except when it is additionally and significantly adjusted, modified or otherwise reworked by a person and only when the reworked work fulfills the requirement of "a sufficient amount of original authorship". The analysis of copyright infringements revealed that the use of copyrighted works without consent in the process of machine learning corresponds to the category of scientific research and is legal according to Directive 2019/790. Such use also complies with the principle of the fair use doctrine in cases where no illegal copy or direct translation is generated, the works are transformed into a new individual product, are used for educational purposes, and do not have a negative impact on the market value of the work. The study also showed that using copyrighted works without permission in the machine learning process may violate copyright when an unauthorized copy or direct translation is generated. Reproduction, distribution, public announcement, rental or sale of an unauthorized copy or translation without the author's consent violates property rights (right to remuneration) and non-property rights (a) using without indicating authorship; b) using the author's honor and reputation in a context that hurts; c) distorting or changing the work in a way that harms the author's honor and reputation). If the author publicly and clearly indicates that he reserves the exclusive right to use the works, the use of these works in art programs does not comply with the permissible purposes provided for in Directive 2019/790 and is considered illegal. The de minimis principle can be applied in general cases when the art program uses the works for educational purposes and transforms it into a new individual result, but when a specific disagreement of the author is indicated or an illegal copy occurs, the exception loses its basis for application. The technological process exception covers only certain parts of the machine learning process, such as temporary copying actions while learning the algorithm, but does not include the long-term creation of databases, providing a copy to users. Data systems in programs are often not only carefully created by carefully selecting their content, but also used in other programs, stored for a long time, etc., which is no longer treated as a short-term technological action. Short-term acts of reproduction also do not include cases of illegal copying, which expand the scope of use of the work and are not a necessary part of the technological process. It was observed, that in order to reduce the risk of copyright violations, it is recommended for software developers to apply a model of providing collective licenses with an extended scope, ensuring the right of revocation of authors against unwanted use. It is also recommended to create databases from works belonging to the category of public art. In order to avoid violations of authorship, programs may name the authors whose works make up the database or generate the result. Authors who do not wish their works to be part of the program's database may publicly and specifically express such objection. If such a work is mistakenly included in the database, the application developers must use their best efforts to ensure the unavailability of these works and prevent their future uploading, and act promptly to remove access to the works or remove them from the websites.