Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Upphovsrättens plats i en tidsålder av generativ AI

Ryberg, Elias LU (2024) LAGF03 20242
Department of Law
Faculty of Law
Abstract
In recent years artificial intelligence has risen in popularity. In relation with the increased interest, we have also seen new applications with said technol-ogy. One of those uses are called “generative AI”. Generative AI is able to create different kinds of works through user made instructions referred to as “prompts”. What was of interest to many was the ability of generative AI like “Midjourney” to create artworks based on those instructions to fit the users request.
But while the technological advancements developed quickly the same could not be said for the legal framework. One problem that arose was the question surrounding how these artworks came to be. These artworks are created by an AI that learns through a process called... (More)
In recent years artificial intelligence has risen in popularity. In relation with the increased interest, we have also seen new applications with said technol-ogy. One of those uses are called “generative AI”. Generative AI is able to create different kinds of works through user made instructions referred to as “prompts”. What was of interest to many was the ability of generative AI like “Midjourney” to create artworks based on those instructions to fit the users request.
But while the technological advancements developed quickly the same could not be said for the legal framework. One problem that arose was the question surrounding how these artworks came to be. These artworks are created by an AI that learns through a process called “text- and datamining” (TDM). This process requires that the AI make digital copies of other works in which it then uses this to learn and extract data from.
The issues with copyright appear when the copies that the AI makes are re-productions from copyrighted works. In this paper this process will be stud-ied and how this process relates to Swedish and EU law.
What will be studied is the so-called DSM directive that regulates the TDM process, in which it will then lead us to article 5 in the infosoc directive that contains a potential exception to copyright. What will be analysed is if gen-erative AI can fulfil the conditions for this exception as well as EU’s view on copyright. This is done to get some insight into the future for generative AI.
What leads to some issues for generative AI is the conditions concerning legal use as well as the fact that the reproduction can’t have any independ-ent economic significance. An analysis concerning different cases where these issues were raised will be done to compare it to generative AI. The conclusion that is reached is that generative AI will have some tall hurdles to climb in order to overcome these conditions because the reproduction that is created within a generative AI will most likely be considered to have inde-pendent economic significance. However, even if this condition should be met it is still unlikely that EU would let generative AI operate like this, con-sidering their rather protective view of copyright and artists ability to create. (Less)
Abstract (Swedish)
Under senare år har artificiell intelligens blivit alltmer populärt. I samband med det ökade intresset har vi sett nya applikationer för denna teknologin. En ny användning blev att man genom viss AI databaser fick fram vad som kallas för ”generativ AI”. Generativ AI kan skapa olika verk på olika alster efter man givit den instruktioner som kallas ”prompts”. Av intresse för många var hur viss generativ AI som ”Midjourney” kunde skapa konstverk utifrån de instruktioner man gav den.
Medan den tekniska utvecklingen fortsatte att utvecklas snabb så kan inte detsamma sägas för den juridiska utvecklingen. Ett problem som uppstod var frågan gällande hur dessa verk faktiskt skapas. Dessa verk skapas genom att en AI lär sig genom en process som... (More)
Under senare år har artificiell intelligens blivit alltmer populärt. I samband med det ökade intresset har vi sett nya applikationer för denna teknologin. En ny användning blev att man genom viss AI databaser fick fram vad som kallas för ”generativ AI”. Generativ AI kan skapa olika verk på olika alster efter man givit den instruktioner som kallas ”prompts”. Av intresse för många var hur viss generativ AI som ”Midjourney” kunde skapa konstverk utifrån de instruktioner man gav den.
Medan den tekniska utvecklingen fortsatte att utvecklas snabb så kan inte detsamma sägas för den juridiska utvecklingen. Ett problem som uppstod var frågan gällande hur dessa verk faktiskt skapas. Dessa verk skapas genom att en AI lär sig genom en process som kallas för ”text- and datamining” (TDM). Denna process involverar att man kopierar andra verk för att sedan utvinna data ur dessa verk som ens AI kan använda och lära sig från.
Det upphovsrättsliga problemet som uppstår är om detta är lagligt om verken som används är upphovsrättsligt skyddade. I denna uppsatsen kommer denna processen att undersökas och hur detta förhåller sig till Sveriges och EU:s lagstiftning.
Det som undersöks är det så kallade DSM direktivet som reglerar TDM pro-cessen, varav detta sedan tar oss sedan till artikel 5 infosoc direktivet för en potentiell inskränkning i upphovsrätten. Det som analyseras är om generativ AI kan uppfylla dessa rekvisit och EU:s värderingar kring upphovsrätten. Detta görs för att försöka få insikt på framtiden för generativ AI.
Det som leder till problem för generativ AI är rekvisiten om laglig använd-ning av ett verk samt om mångfaldigandet har självständigt ekonomisk be-tydelse. Varav en analys görs utifrån andra fall där dessa rekvisit var av be-tydelse och detta jämförs sedan med hur detta förhåller sig till generativ AI. Slutsatsen blir att generativ AI kommer ha stora svårigheter att uppfylla dessa rekvisit eftersom mångfaldigandet som skapas i en AI kan troligtvis bedömas ha självständigt ekonomisk betydelse. Men även om detta rekvisit skulle vara uppfyllt så vore det inte sannolikt att EU skulle tillåta generativ AI att fortsätta att existera på ett sådant oreglerat sätt med tanke på hur de-ras syn det kreativa skapandet och upphovsrätten som helhet. (Less)
Please use this url to cite or link to this publication:
author
Ryberg, Elias LU
supervisor
organization
course
LAGF03 20242
year
type
M2 - Bachelor Degree
subject
keywords
EU-rätt, immaterialrätt, upphovsrätt, generativ AI
language
English
id
9179548
date added to LUP
2025-03-20 14:19:36
date last changed
2025-03-20 14:19:36
@misc{9179548,
  abstract     = {{In recent years artificial intelligence has risen in popularity. In relation with the increased interest, we have also seen new applications with said technol-ogy. One of those uses are called “generative AI”. Generative AI is able to create different kinds of works through user made instructions referred to as “prompts”. What was of interest to many was the ability of generative AI like “Midjourney” to create artworks based on those instructions to fit the users request. 
But while the technological advancements developed quickly the same could not be said for the legal framework. One problem that arose was the question surrounding how these artworks came to be. These artworks are created by an AI that learns through a process called “text- and datamining” (TDM). This process requires that the AI make digital copies of other works in which it then uses this to learn and extract data from. 
The issues with copyright appear when the copies that the AI makes are re-productions from copyrighted works. In this paper this process will be stud-ied and how this process relates to Swedish and EU law. 
What will be studied is the so-called DSM directive that regulates the TDM process, in which it will then lead us to article 5 in the infosoc directive that contains a potential exception to copyright. What will be analysed is if gen-erative AI can fulfil the conditions for this exception as well as EU’s view on copyright. This is done to get some insight into the future for generative AI.
What leads to some issues for generative AI is the conditions concerning legal use as well as the fact that the reproduction can’t have any independ-ent economic significance. An analysis concerning different cases where these issues were raised will be done to compare it to generative AI. The conclusion that is reached is that generative AI will have some tall hurdles to climb in order to overcome these conditions because the reproduction that is created within a generative AI will most likely be considered to have inde-pendent economic significance. However, even if this condition should be met it is still unlikely that EU would let generative AI operate like this, con-sidering their rather protective view of copyright and artists ability to create.}},
  author       = {{Ryberg, Elias}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Upphovsrättens plats i en tidsålder av generativ AI}},
  year         = {{2024}},
}