Researchers have pioneered a selective unlearning technique for medical AI systems that enables the removal of sensitive patient data when consent is withdrawn, without the need to retrain the entire AI model. This approach allows the AI to forget specific information considered legally or ethically risky while maintaining its diagnostic performance and clinical expertise. Traditional methods to delete personal data require costly and time-consuming retraining of models on remaining datasets, which can degrade overall accuracy. The new method achieves targeted unlearning, selectively erasing specific patient-related knowledge without impacting the AI’s broader clinical capabilities. This innovation helps healthcare providers comply with strict patient privacy regulations and data removal requests, mitigating legal risks around data usage. The technique may be especially valuable as medical AI adoption grows and concerns about data privacy intensify. By preserving the AI’s core skills, this method balances the ethical imperative to respect patient rights with the practical need to retain high-quality diagnostic tools. Recent research developments detail frameworks like “Erase to Retain,” which guide controlled unlearning through low-rank adaptation, ensuring that the AI discards only the designated sensitive information. Overall, this advancement marks an important step toward more responsible and patient-centric application of AI in medicine, promoting trust and regulatory compliance without sacrificing innovation or accuracy in clinical decision-making.