There is no dermatology panel at FDA.

Skin devices are reviewed under General & Plastic Surgery, dermatopathology under Pathology, some imaging under Radiology. Which means FDA’s own list of AI-enabled devices cannot be filtered for dermatology at all. The cohort below had to be assembled by hand, one product code and one regulation at a time.

Two things became visible once it was. FDA itself describes only two of these ten devices as using artificial intelligence or machine learning — the rest it calls software algorithms, classifiers, automated systems. And only one of the ten has published performance broken out by skin tone. Two more report the Fitzpatrick makeup of their study population without stratifying results by it. Seven say nothing about skin tone at all.

Every field here comes from an FDA primary document: a clearance letter, a De Novo granting order, a PMA approval order, a summary of safety and effectiveness data. Where FDA does not state something, the table says not stated rather than leaving it blank. That distinction carries more than it looks — a blank cell reads as “no,” and these are not the same claim.