Derm Dispatch: Navigating the Ethics of AI-Assisted Dermatology

In the first Derm Dispatch episode of 2026, host Renata Block, DMSc, MMS, PA-C, spoke with emergency physician Harvey Castro, MD, MBA, nicknamed “Dr GPT,” about the expanding role of artificial intelligence (AI) in dermatology and the broader clinical enterprise. Castro framed AI not as a replacement for clinicians but as a permanent and necessary clinical partner—one that should be budgeted like electricity or staffing rather than justified solely by short-term return on investment.

A major barrier to adoption, Castro noted, is the technical and financial burden of integrating AI into electronic medical records (EMRs). However, he argued that delaying adoption until a clear ROI is demonstrated misses the reality that AI will soon be embedded in standard care. Instead, institutions should work backward from their clinical goals, such as improving efficiency, detecting disease earlier, and reducing morbidity, and select AI tools that align with those outcomes.

Equity and bias were central themes of Block and Castro’s conversation. He emphasized that AI systems are only as good as the data used to train them, pointing out that dermatologic training materials historically underrepresent skin of color. If these biased datasets are used to train AI, the technology risks perpetuating diagnostic inequities, particularly for Fitzpatrick skin types V and VI. To address this, he advocated for localized training of smaller AI models using institution-specific data and for federated learning systems, in which anonymized data “weights” are shared across sites to improve performance without compromising patient privacy.

Trust and transparency, Castro argued, will be critical for patient acceptance of AI-assisted dermatology. He proposed an “AI Bill of Rights” that clearly discloses how AI tools are used, their known biases, and their limitations—similar to a nutrition label—so patients understand how AI informs their care.

Castro also warned against over-automation. While AI can enhance diagnostic accuracy and efficiency, clinicians must retain foundational diagnostic skills and the ability to challenge machine output. He called for continued emphasis on clinical reasoning in training, including deliberate periods when AI tools are turned off to preserve clinical gestalt.

Looking ahead, Castro described a future in which dermatology moves beyond episodic diagnosis toward longitudinal, predictive care. Digital twins, genetic data, wearables, and AI-driven analytics could enable earlier disease detection, personalized treatment trajectories, and even preventive interventions before pathology becomes clinically evident, reshaping how chronic inflammatory diseases and skin cancer are managed over a patient’s lifetime.