If you’ve tried to count how many AI dermatology devices the FDA has actually cleared, you’ve probably hit a wall: the data doesn’t filter cleanly by specialty. That isn’t a quirk of how the data is organized. It reflects a real structural gap in how skin devices are regulated.
Dermatology Has No Panel
Skin devices don’t have their own FDA review panel. They’re classified under General & Plastic Surgery (21 CFR Part 878), the same regulatory bucket as surgical instruments, with no dermatology-specific category.
In practice, there’s no way to pull “every FDA-authorized AI dermatology device” as a clean query against the FDA’s own database. Anyone trying to map this space has to build that list by hand.
Two Categories Take Shape
The FDA is building a more specific framework within that broader category. Two device classifications now exist:
- 21 CFR 878.1830: software-aided diagnosis of skin lesions suspicious for cancer.
- 21 CFR 878.1840: software-aided skin wound assessment.
DermaSensor (DEN230008) created the first category. In January 2024, it became the first AI-enabled device authorized to evaluate skin lesions for cancer in the primary care setting, meaning non-dermatologists can use it, not just specialists.
The DermaSensor Case Study
DermaSensor is worth studying closely because it has the most publicly disclosed skin-tone performance data in the category. Its premarket pivotal studies, DERM-ASSESS III and DERM-SUCCESS, showed strong overall sensitivity. But the large majority of enrolled patients had lighter Fitzpatrick skin types.
Recognizing that gap, the FDA required a postmarket validation study specifically enrolling patients with Fitzpatrick skin types IV, V, and VI, to confirm the device performs consistently in those populations.
The company has since reported that expanded studies show consistent performance across all six Fitzpatrick types. That is the manufacturer’s own reporting, not an independent third-party audit.
This pattern, strong premarket data with a real skin-tone gap addressed through a required postmarket study rather than blocked at clearance, is likely to become a template that other AI dermatology devices follow or are measured against.
The Regulatory White Space
As of this writing, there’s no FDA authorization for AI-enabled tools in several areas companies are actively building for:
- Acne: severity grading or treatment planning.
- Hair loss: alopecia assessment.
- Teledermatology: triage of patient-submitted images.
- Dermatopathology: whole-slide image analysis.
- Aesthetics: energy-based device treatment planning.
Each is open ground, with no established pathway and no precedent device to benchmark against. That’s a barrier or an opportunity, depending on where you sit.
The Bigger Picture
For clinicians, “FDA-cleared AI dermatology device” describes a much narrower category than the marketing around AI in dermatology suggests. Right now, it largely means DermaSensor and the lesion and wound assessment framework building around it.
For companies, the postmarket skin-tone study is worth planning for from the start, not treating as a late-stage compliance problem. Performance across all skin tones is a patient safety question, and regulators are starting to treat it that way.
