AI medical tools need stricter checks to protect patients, safety commissioner says
Post-market surveillance of AI health tools must be “beefed up” to protect doctors as well as patients, England’s patient safety commissioner says.
Henrietta Hughes also told The BMJ it was vital to establish clarity on where clinical liability sits when, not if, AI tools harm patients.
Hughes, a GP and a former medical director at NHS England, is deputy chair of the National Commission into the Regulation of AI in Healthcare. The commission was set up by the Medicines and Healthcare Products Regulatory Agency (MHRA) to help guide development of a new regulatory framework for AI medical devices.
The commission published interim findings from its consultation and engagement process last week.
Hughes said some clear themes had already emerged during the process of engagement with patients, the public, and doctors. Among the most pressing was the need for greater surveillance of AI tools after approval, so the MHRA can act if patients are at risk.
Hughes told The BMJ, “It’s really important that real time, real life monitoring happens when a device like AI is deployed in a real life clinical environment, particularly if the population of patients may be different from the population used to feed the model.”
Hughes added that while medicines have to pass an “extremely high hurdle” and evidence base to reach the market, AI—where new products are rapidly launched and updated—is different.
“We know that AI can change once it’s actually deployed, and so it’s important that the regulations are able to be updated to take account of that and to ensure that all medical devices, and particularly AI, are safe across its whole life cycle,” she said.
“Whether we’re using the yellow card system or other kinds of ‘always-on’ postmarket surveillance and postmarket monitoring, that side of things really needs to be significantly beefed up if we’re going to lower the hurdles for products to come onto the market.”
Source: BMJ, 18 June 2026