A new review published in Nature warns that heavy reliance on AI can cause “deskilling,” where professionals gradually lose abilities they built over years. The article says AI may boost performance in the short term, but over time it can weaken independent work, from doctors missing medical findings to programmers understanding less of the code they produce.
The concern is already widespread in healthcare. A U.S. survey found that 70% of nurses and 77% of physicians worry that overdependence on AI will damage their professional competence. One study in The Lancet Gastroenterology & Hepatology followed experienced colonoscopy specialists who began using real-time AI to detect precancerous lesions. When the system was not available, their detection rate fell by about 21%.
The researchers said prolonged use may reduce doctors’ concentration, motivation, and sense of responsibility when working without AI. Study author Dr. Yuichi Mori of the University of Oslo said more research is needed to confirm how large the problem is, but warned that users should understand they may lose part of their skills. He said there is still no proven way to prevent deskilling and expects it to become a major research topic in the coming decade.
A separate Anthropic study looked at 52 software engineers. Half used an AI assistant during a coding task and were then tested on what they had learned. Those who used AI scored an average of 50%, compared with 67% for those who did not. The gap was especially strong on questions requiring error detection and code understanding, suggesting the AI users completed the task but absorbed less of the underlying logic. The researchers said the effect is especially worrying for students and early-career programmers, because AI can help them produce better results while weakening their learning and skill development. They recommend continuing to practice without AI, understanding how the systems work, and not accepting their answers uncritically.
