Right to Erasure in Generative Artificial Intelligence Models: Functional Neutralization Standard

Ramos-Zaga F.A. · Revista de la Facultad de Derecho y Ciencias Politicas (2025) · DOI: 10.18566/rfdcp.v55n143.a7

Resumen

The training of generative artificial intelligence models has reshaped the challenges facing data protection by creating systems in which the removal of personal information is technically difficult once it has been incorporated into the model. Against this background, this article examines the legal scope of the right to erasure under Article 17 of the General Data Protection Regulation (GDPR) when personal data have been used to train a generative artificial intelligence model. The study adopts a qualitative documentary approach grounded in doctrinal legal study and an interpretive research paradigm. The analysis demonstrates that the technical infeasibility of physically removing embedded personal data does not constitute an independent exception to the right to erasure. To address this issue, the article develops a verifiable functional neutralization standard based on a three-part test comprising non-reproducibility, non-inferability, and proportionality of the required technical effort. It concludes that the obligation to erase personal data remains mandatory, while functional neutralization is compatible with the right to erasure only where it produces verifiable outcomes, effectively reduces the risk of identification, and is supported by appropriate documentation, auditing, and oversight.

Instituciones: Universidad Privada del Norte