Innovation or Infringement?

Sammanfattning

This thesis investigates the legal and ethical implications of generative artificial intelligence (AI) within the European Union copyright framework, with a particular focus on text and data mining (TDM) practices. It explores how AI systems are trained using vast datasets - often scraped from online content - raising concerns around authorship, consent, and the preservation of attribution. The study critically examines the EU Copyright Directive 2019/790 and its TDM exceptions under Articles 3 and 4, assessing their adequacy in regulating AI development. Through doctrinal analysis and academic commentary, it identifies significant gaps in current legislation, particularly around data provenance and the enforceability of moral rights such as attribution and integrity. The thesis further evaluates the feasibility of applying existing copyright doctrines to machine-generated content and reviews proposals advocating for enhanced transparency and governance. Special attention is given to the emerging EU AI Act (2024) copyright law. The Act primarily focuses on the input and training stage of generative AI. Ultimately, the research advocates for a harmonised legal framework that balances innovation with the rights of human creators, suggesting that attribution and licensing must be embedded into the technical and legal architecture of AI development.