A study published in The Lancet Gastroenterology & Hepatology reports that routine reliance on artificial-intelligence software during colonoscopies can erode physicians’ skills when the technology is absent. Researchers analysed performance at four endoscopy centres in Poland participating in the ACCEPT trial, which alternated procedures with and without an AI polyp-detection system. Across 1,443 diagnostic colonoscopies conducted between September 2021 and March 2022, the adenoma detection rate in standard, non-AI-assisted exams dropped to 22.4% after three months of exposure to the algorithm, compared with 28.4% before its introduction—a six-percentage-point, or roughly 20%, decline. Multivariable analysis identified prior AI exposure as an independent factor lowering detection odds (odds ratio 0.69). The authors say the findings provide early evidence of a potential “deskilling” effect from clinical AI and underscore the need for safeguards, such as periodic unassisted practice and complementary training, as hospitals accelerate deployment of automation across medical specialties.