Ethical considerations in generative artificial intelligence for synthetic data generation in ophthalmology

Generative artificial intelligence (AI) is beginning to reshape medical research, with ophthalmology at the forefront of this transformation. This narrative review examines the ethical considerations and challenges of using generative AI - particularly Generative Adversarial Networks (GANs) and diffusion models - to create synthetic data for ophthalmic research. Synthetic data offers genuine solutions to long-standing obstacles: data scarcity, augmentation of datasets for rare diseases, and privacy-preserving multi-institutional collaboration. Yet the same capabilities introduce formidable risks. Key concerns include patient re-identification from ostensibly anonymous synthetic data, the perpetuation and amplification of algorithmic biases inherent in source datasets, the prospect of scientific misconduct through "deepfake" medical images, and unresolved questions of data ownership, governance, and legal liability. Clinical validity cannot be inferred from computational metrics alone; it requires a multi-faceted evaluation framework that incorporates expert clinical review and downstream task performance. To harness the potential of generative AI while mitigating its risks, a proactive, interdisciplinary approach is essential - combining rigorous technical validation with ethical oversight and clear regulatory guidance. This article concludes with policy recommendations to foster responsible innovation and ensure that this powerful technology is developed and deployed safely, equitably, and aligned with the core tenets of scientific integrity and patient care.