From Algorithmic Destruction to Algorithmic Imprint: Generative AI and Privacy Risks Linked to Potential Traces of Personal Data in Trained Models