Artificial intelligence is moving through medicine faster than regulation, ethics, and professional reflection can keep pace, according to a correspondence published in The Lancet. The authors argue that the prevailing assumption — that if a technology can be developed, it should be developed, with its place in clinical care debated afterwards — is backwards. They believe that sequence should be reversed.
The correspondence, titled «Regulate prospective development of medical AI now», calls for regulatory frameworks to be established before medical AI tools are widely deployed, rather than retrofitted after they are already embedded in clinical practice. The authors frame the issue as a matter of sequencing: oversight should come first, not last.
The concern is not hypothetical. Medical AI is already being applied across a range of clinical tasks, from diagnostic imaging and pathology to risk prediction and treatment planning. Each of these applications carries direct consequences for patient safety, and each raises questions about accountability when an algorithm contributes to a clinical decision. The correspondence suggests that current governance structures have not kept up with the speed of deployment.
The authors point to a gap between technological capability and the ethical and professional frameworks meant to guide its use. Regulation, they write, has struggled to match the pace of AI development in medicine. The same is true of professional reflection — the slower, deliberative work of considering what role a technology should play in patient care before it becomes routine.
Their proposed reversal is significant because it challenges a common pattern in health technology adoption. Typically, a tool is developed, validated in studies, and then introduced into practice, with policy and ethical debate following. The correspondence argues that for medical AI, this order exposes patients and clinicians to risks that could be anticipated and addressed earlier.
The piece does not propose a specific regulatory mechanism, but its central claim is clear: prospective regulation — rules and oversight designed before or alongside development — should replace the reactive approach that has dominated so far. That would mean involving regulators, ethicists, and clinicians at earlier stages of the development pipeline, rather than treating them as downstream reviewers.
The correspondence also implies a broader question about how medical innovation is governed. If the default assumption is that any developable technology will be developed, then the burden falls on regulators and professional bodies to catch up. Reversing that assumption would place the burden on developers and institutions to demonstrate, in advance, that a tool is appropriate for clinical use and that its risks are understood.
For clinicians, the practical stakes are immediate. AI tools that influence diagnosis or treatment can affect patient outcomes, and when something goes wrong, it is not always clear who is responsible — the developer, the institution, or the clinician who relied on the output. Prospective regulation could clarify those lines of accountability before harm occurs.
The correspondence appears in The Lancet as a contribution to the ongoing debate over how medicine should absorb AI. It adds a clear position to that debate: that the current order of development first and regulation later is a mistake, and that the sequence should be reversed now, while medical AI is still taking shape.
15
