Verifying Human Works: A Comparative Analysis of the Right of Attribution in relation to AI Generated Media

Sammanfattning

With the rise of AI generated media, a shift in the audience perception of AI generated content has taken place, where studies show that content which is labelled as AI will be less interacted with compared to human labelled content. The thesis therefore examines possible methods for verifying human made content, in particular, whether copyright’s right of attribution can act as a veri-fier for human works within the EU. The initial chapter examines prevailing regulation regarding AI generated content in view of distinguishment, where the AI Act has been examined. It concludes that there are obligations on providers and deployers for labelling AI generated content. The limitations for deployers to only mark deepfakes and in a way it does not impede their artistic freedom is emphasized as the greatest implication for the audience to distinguish between human made works and AI generated media. Since the labelling obligations under the AI Act are not absolute, the thesis examines the moral right of attribution, under the Berne Convention and with-in the EU. The right of attribution, as well as other moral rights, are not harmonised within the EU, where the main reasoning has been due to a lack of effect on the internal market. The opposite is however concluded in view of the circulation of works on the internet as well as that some harmonisation has been done through legislation and the CJEU’s jurisprudence. It is further concluded that in order to attain a smooth functioning of the internal market as well as upholding the interest of authors, a harmonisation of the right of attribution is necessary. In what way the EU is to construe the right of attribution in relation to human verification, Sweden and France’s right of attribution are examined. Where the findings are that the right is similar in some aspects, as in its personal nature. However the right has a stronger standing in France due to its perpetuity and more extensive applicability compared to Swedish law. The similarities and differences in the two legal systems are analysed regarding human verification, where the French system seems to be more favourable in the majority of cases. The thesis concludes that the obligations under the AI Act is not a complete measure in order for the audience to distinguish AI generated media, where the alternative is the right of attribution. The EU can legislate that AI markings are not be confused with authors’ attributive marks as well as a copyright symbol similar to the Commissions proposed for AI generated media. Questions about registrations of copyright is however scrutinized in view of author’s right to remain anonymous.