Responsible AI for Skin Health

1. Evidence before claims

Any performance claim we publish has to be traceable to a documented evaluation. Our reported figures are 95.3% concordance with board-certified dermatologists in independent clinical validation, and 96.48% benchmark test accuracy on public datasets including HAM10000 and DermNet. Benchmark accuracy and real-world concordance are different measures, and we label them differently on purpose.

2. Human expertise matters

AI can handle the scalable, repeatable first interaction: an image, a question, a comparison over time. Clinical judgement, physical examination and diagnosis remain the responsibility of healthcare professionals. Our clinical review network supports our medical content, review services and quality processes.

AI where it can scale. Human expertise where it matters.

3. Inclusive performance

Dermatology AI has a well-documented history of underperforming on darker skin, largely because public training data has been skewed towards lighter skin types. ScanSkinAI is designed and evaluated across all six Fitzpatrick skin types, and skin-tone performance is part of our ongoing evaluation rather than a one-off check. Where evidence is weaker for a given presentation, the right answer is to say so and to route the person towards professional review.

4. Privacy by design

Skin images are personal health information and are treated that way: encrypted transfer and storage, access controls, defined retention periods, and the ability for users to delete their account and associated data. Our security and quality systems follow ISO/IEC 27001 and ISO 13485. In workplace and community programmes, individual results belong to the individual — organisers see participation, not personal health information.

5. Clear limitations

ScanSkinAI is a screening and wellness-support tool. It does not replace:

• dermatologist examination
• dermoscopy or physical examination
• biopsy
• laboratory testing
• medical diagnosis
• treatment planning

A screening result — including a low-concern result — is never a reason to ignore a change that is worrying you. ScanSkinAI is registered with the MHRA as UKCA Class I medical device software (ref 2025111201444906).

6. Appropriate escalation

A screening tool is only responsible if it tells people what to do next. Every result includes guidance on whether professional review may be appropriate, and higher-concern results carry a clear prompt to seek medical assessment. Where users want a human opinion sooner, an optional dermatologist review pathway is available.