Using ambient AI in clinical consultations: reframing policy around clinical audit and patient safety

Recent work has mapped what is legally permitted, commercially practiced, and technically coming for ambient AI in clinical care, yet it has not addressed the prior question of what the recording is for. This Perspective offers a policy argument, not an empirical evaluation. Documentation, workflow efficiency, patient recall, medico-legal transparency, and quality improvement are all legitimate purposes of consultation recording; we propose that one further, high-value purpose has been under-recognized: clinical audit. The high-fidelity transcript makes communicated clinical reasoning observable at scale, aligning consultation recording with the WHO Global Patient Safety Action Plan 2021-2030. Examining the regulatory, vendor, patient, and clinician perspectives, we argue that European law permits this purpose but does not supply it, and that consent should be reframed around patient protection rather than institutional convenience, a protective framing that is itself a hypothesis to be tested. Vendors, in turn, must redesign their products to be audit-ready, with transcription accuracy reported transparently across languages and populations. The efficiency gains that will drive adoption and the audit purpose that could convert adoption into patient value draw on the same artifact and need not compete. As a testable policy proposal, we call for a minimum audit-ready transcript standard, piloted in parallel across high-income and low- and middle-income settings, so that the clinical value of ambient AI is built in from the start rather than retrofitted after high-income adoption.