MANCHESTER, England – AI is already helping to triage patients with suspected skin cancer in parts of the NHS. Now, dermatologists are turning their attention to a different challenge: ensuring the technology is backed by robust evidence, clear standards, and effective oversight before it becomes more widely embedded in routine care.

The Skin and Responsible AI (SkinRAI) Consortium is a new clinician-led initiative set up by the British Association of Dermatologists and launched here during the association’s 2026 annual meeting. 

Rubeta Matin, MBBS, PhD, cofounder and clinical director of SkinRAI, said the consortium’s goal is to set the standards for how AI should be developed, evaluated, and monitored in dermatology, as innovation is often moving faster than the evidence needed to support it.

“The evidence, standards, digital infrastructure, and assurance methods to support the safe use of AI remain quite immature. In order to scale up, we really need to get in front of this,” said Matin, who is an associate professor and consultant dermatologist for Oxford University Hospitals NHS Foundation Trust.

One challenge is that many AI tools have yet to be independently evaluated across different healthcare settings.

AvailableAI technologies have so far been developed on datasets derived from single populations, pathways, or services, so it cannot be automatically assumed that they will work in a different situation. 

To address this, Matin said, “We’re looking to develop national data assets so that we can actually support [the development of AI] technologies going forward.” 

These assets would be federated, meaning that while data will be held by the institutions that have collated it, they will be accessible nationally to anyone that wants to use them. 

Matin outlined four main functions for SkinRAI: developing trusted approaches to data and evidence generation; establishing evaluation methods for emerging AI technologies; identifying evidence gaps and examples of good practice; and supporting clinicians through practical guidance and implementation frameworks.

She stressed that SkinRAI will not promote specific AI products, tell clinicians which systems to use, develop AI technologies itself, or act as a regulator.

Regulating AI Use 

The question is no longer whether AI should be used in healthcare, but under what conditions it can be trusted, said Alastair Denniston, MB BChir, PhD, chair of the National Commission into the Regulation of AI in Healthcare.

The non-statutory advisory body he chairs, established by the Medicines and Healthcare products Regulatory Agency, is due to report later this year with recommendations on how AI should be regulated in healthcare.

Denniston, who is professor of regulatory science and innovation at the University of Birmingham and an honorary consultant ophthalmologist at University Hospitals Birmingham NHS Foundation Trust, observed, “Our ability to deliver high-quality care has improved, yet I’m excited to see what we can do in 10 years’ time and I think AI is part of that story.”

AI-Based Skin Screening 

Dermatology in the NHS is already a trailblazer in its adoption of AI, particularly for skin cancer assessment, where image-based algorithms are increasingly being evaluated to support referral decisions.

One such tool is DERM (Deep Ensemble for Recognition of Malignancy), a class III CE-marked AI medical device that assesses dermatoscopic images of skin lesions for skin cancer.

DERM is already in use in some NHS practices under conditional approval from the National Institute for Health and Care Excellence to act autonomously to help triage people with suspected skin lesions into, or away from, specialist services.

One year into a 3-year evidence-generation period, early real-world data are beginning to show how the technology performs in routine clinical practice and whether it can help ease pressure on overstretched dermatology services. 

Charles Earnshaw, MB BChir, PhD, a clinical lecturer at the University of Manchester and specialist registrar in dermatology for the Northern Care Alliance NHS Foundation Trust, presented audit data on the diagnostic performance of DERM in the urgent skin cancer screening pathway.

The findings, which were also published recently in Skin Health and Disease, showed that in its first 2 months of use at the trust, DERM enabled 274 of 1230 patients (22.3%) to be discharged before seeing a dermatologist.

“In our trust, we still have independent dermatologists review the image before discharge” or referral into the centre, Earnshaw said. 

The AI tool was highly sensitive, detecting 95% of cancers compared with 88.5% for dermatologists. However, dermatologists remained more specific, at 62.1% vs 46.5%.

Based on final pathology, the AI tool was also less accurate than dermatologists at identifying the correct diagnosis (28.6% vs 61.6%). For comparison, Earnshaw noted that GPs are accurate in 22.5% of cases.

It correctly identified the tumour or lesion type in 51.4% of cases, compared with 75.5% for dermatologists.

The findings also underscored the importance of human oversight. Four cancers were diagnosed in lesions that the AI had deemed benign and would have been discharged had no independent human review of the results been part of the process. 

Experience from routine NHS practice painted a similarly encouraging, but cautious, picture. Alexandra Kemp, a consultant dermatologist working in private and NHS practice, reported on the implementation of DERM at an NHS district hospital in Amersham, England.

Like much of the UK, “[w]e’ve got rising demand and lack of capacity,” Kemp said. Other means such as teledermatology solutions have been looked at but “we just don’t have the humans to read those photos, so for us we had to look elsewhere.”

Since starting to use the AI tool a couple of years ago as part of a pilot, Kemp reported that 4723 referrals had been assessed up to June 23 of this year. Of these, 901 were automatically referred for an urgent face-to-face consultation. After review by a dermatologist, 31% were discharged without needing to attend.

The remaining 3822 referrals underwent independent dermatologist review, with 27% discharged without a face-to-face consultation. Just 14% of cases checked through a second review of the AI result were overturned. 

“We’re not aware of any missed invasive melanomas,” said Kemp. The system did miss two low-risk squamous cell carcinomas, three basal cell carcinomas, and one case of melanoma in situ that “shouldn’t have gone through the pathway. It was too large for the dermatoscope, so was seen back anyway.”

Kemp said, “As a small dermatology unit, this has had a massive impact, and it’s the first time that we’ve had something that’s really helped us.” 

Even so, Kemp cautioned that more independent evaluation is needed.

“I do think we need more analysis of the impact of the AI in the NHS in a real-world setting, which should be independent of the data from the AI company.” 

Matin, Denniston, Earnshaw, and Kemp reported no relevant financial relationships. DERM is manufactured by Skin Analytics. 

Sara Freeman, BSc, MSc, is a freelance medical journalist based in London, UK. She has been reporting for specialist healthcare news organisations for more than 20 years.