Why the generative AI models do not like the right to be forgotten: a study of proportionality of identified limitations

Abstract

The article explores the limitation of one of the privacy and data protection rights when using generative AI models. The identified limitation is assessed from the perspective of the ‘essence’ of the right to the protection of personal data. With the further aim of assessing the limitation, the author explores whether the right to be forgotten (RTBF) is relevant or effective in an AI/machine learning context. These considerations are focused on the technical problems encountered when applying the strict interpretation of the RTBF. In particular, the antagonism between, on the one hand, the values of privacy and data protection rights, and on the other, the technical capabilities of the producer of the generative AI models, is further analysed in this context. As the conclusion emphasizes that the RTBF cannot be practicably or effectively exercised in the machine learning models, further considerations of this exposed limitation are presented. The proportionality principle, as an instrument that supports the proper application if there is any limitation of the conflicting rights, has been utilized to depict the qualitative approach. The integration of this principle supports the conclusion by identifying a more efficient way to address some regulatory issues. Hence, the conclusion of the article presents some suggested solutions as to the interpretation of this right in the light of this new technological advancement. Ultimately, the paper aims to address the legal conundrum of how to balance the conflict between the interest of innovative use of the data (the data producer’s right) and privacy and data protection rights.

Keywords

• the right to be forgotten
• the data producer’s right
• the essence of fundamental rights
• proportionality
• AI Act
• Data Act