Flawed AI in Healthcare: Civio Wins European Press Prize for Skin Cancer Algorithm Investigation

The Spanish non-profit organisation Civio has won the European Press Prize 2026 in the "Innovation" category. The award, which includes a €10,000 prize, was presented in Lisbon. The winning investigation focuses on an artificial intelligence system used for diagnosing skin cancer.

Journalists Ángela Bernardo, María Álvarez del Vayo, Carmen Torrecillas, and Adrián Maqueda investigated the "Quantus Skin" algorithm, which the Basque Country health service, Osakidetza, began introducing into primary care centres. The tool was intended to help general practitioners distinguish benign moles from potentially malignant lesions and speed up patient referrals to dermatologists.

In 2022, Osakidetza awarded a €1.6 million contract to the company Transmural Biotech to implement AI algorithms for analysing medical images, including photographs of suspicious skin lesions. The contract required the algorithm to correctly identify cancer in at least 85% of cases. In practice, however, it succeeded only 69% of the time, missing every third melanoma. Additionally, the system misidentified one in five benign moles as melanoma, risking unnecessary specialist referrals for roughly 20% of examined patients.

The investigation also revealed that Quantus Skin was trained almost exclusively on images of patients with white skin. It relied initially on a database of over 56,000 images from the international ISIC repository, collected predominantly in Western hospitals, and subsequently on images from 513 patients at the Ramón y Cajal Hospital, all of whom were Caucasian.

"I don't want to get into the issue of ethnic minorities and all that, because this tool is used by the Basque Country, by Osakidetza. I provide the tool with its limitations," David Fernández Rodríguez, CEO of Transmural Biotech, told journalists.

The Basque Government stated that it did not consider it necessary to take measures "to ensure equality and non-discrimination."

"For skin cancer diagnosis, a sensitivity rate of 70% is very bad. It is an extremely low figure. If you let someone take a picture to tell you if it might be a melanoma, and they get it wrong in one out of three cases, such a test is unacceptable for primary skin cancer screening. We must demand more," explains Dr Josep Malvehy Guilera, Director of the Skin Cancer Unit at Hospital Clínic in Barcelona.

Dr Adewole Adamson, a professor of dermatology at the University of Texas, who warned about the discriminatory consequences of AI back in 2018, summarised: "Biased algorithms deprive patients of the potential benefits of this revolutionary technology. The problem is not the algorithm itself, but how it is developed."

The European Press Prize jury called Civio's investigation "an exemplary piece of data journalism with a clear methodology" and emphasised its undeniable public significance.