The AI radiologist found everything the human missed. Then the lawsuit started, and the question everyone had been sidestepping finally arrived in a courtroom: when a machine is objectively better at diagnosis, who carries the liability when someone chooses to override it?
A London teaching hospital deployed a chest X-ray model in early 2025 that detected 23% more actionable abnormalities than its senior radiology team in a six-month controlled study. The numbers were clean, the validation robust. Then a patient whose follow-up was delayed because a radiologist disagreed with the model’s finding developed stage III lung cancer. The suit names both the hospital trust and the model’s developer — arguing that deploying a system known to be more accurate, while retaining a human override process that introduced error, created a duty of care that neither party fulfilled.
The legal logic is still untested, but the insurance industry is already moving. Two major medical malpractice insurers announced premium tiers in 2026 that differentiate between practices that use AI as a “consult” versus those that use it as a “co-signer.” The technologists built a better diagnostic tool. The lawyers are now building the framework for who pays when it matters.
