Taking notes or something more?
The potential of AI has definitely not been lost on the healthcare sector. AI has the ability to solve many long-standing problems in modern healthcare: lack of resources, risks of human error and difficulties in serving a rapidly evolving patient base. While risks and concerns exist, the AI transformation of the patient journey and other healthcare processes seems inevitable, both in the private and public sectors.
AI scribes, namely AI tools with the ability to convert human interactions into transcripts, visit summaries and other clinical documentation, appear to be a key example of this development. The typical use case is that an appointment between a patient and medical professional is recorded by an AI tool, converted into text and arranged into legally required medical record entries. As a result, the doctor can focus fully on the patient interaction without having to take meticulous notes and save time afterwards as the AI tool has already drafted the required entries.
Legally, things get interesting when we look at what the AI scribe actually does and the extent to which the medical record entries and subsequent clinical activities rely on the AI scribe’s output.
Hitting the right class under the MDR and AI Act
The typical issue with any new technology is how it interacts with legal classifications in existing legislation. For AI scribes, a particularly debated issue is whether they fall under the EU Medical Device Regulation (MDR) and, if so, under what classification. This, in turn, determines what risk class the AI scribe falls under in accordance with the EU Artificial Intelligence Act (AI Act) – namely whether it should be considered a high-risk AI system, to which the majority of the AI Act’s substantive requirements apply.
Under the AI Act, an AI medical tool is considered high risk if it meets both of the following conditions1:
- the tool is a medical device under the MDR or a safety component thereof; and
- the medical device is subject to a third-party conformity assessment by a notified body in accordance with the MDR.
There has been considerable variation in assessments of whether AI scribes meet these criteria. In some cases, especially where the AI scribe merely provides a first draft of a patient entry for the doctor to review and confirm, the understanding has been that the AI tool might not even constitute a medical device. However, with the development of AI scribes and authority positions, there seems to be a push toward generally interpreting AI scribes as medical devices and even medical devices impacting patient care to a degree that requires classification subject to third-party conformity assessment.
Debate continues, and healthcare operators and product manufacturers are advised to monitor relevant authority interpretations in this area. The recent regulatory activity of the Swedish Medical Products Agency (Läkemedelsverket) has been considered particularly relevant. Ultimately, the careful design and determination of relevant clinical processes (i.e., how the outputs of the AI scribe are applied in practice) will be decisive.
Missed opportunity with the Digital Omnibus?
For AI falling under both AI Act high-risk and MDR regimes, the continued concern of the AI medical device sector has related to the risk of over-regulation.
In terms of regulatory simplification, a key effort was the finalisation of the so-called Digital Omnibus on AI Regulation this year, which brought deadline extensions and certain reliefs to the AI Act. As part of this process, there had been considerable pressure for medical devices to be excluded from the direct application of the AI Act.
The medical device sector was disappointed to learn that it was not granted such an exclusion as proposed. As a result, relevant AI medical devices require compliance efforts under both the MDR and AI Act. However, such systems benefit from the AI Act’s deadline push granted under the Digital Omnibus. Moreover, the European Commission was empowered to adopt delegated acts to limit the application of AI Act requirements where reasonable to avoid regulatory overlaps, including regarding medical devices, so the final regulatory position is not yet settled.
Our team is happy to advise on any questions you may have about AI and healthcare regulation in Finland.
1 An AI medical tool may also be considered high risk if it is intended to be used by public authorities or on behalf of public authorities to evaluate the eligibility of natural persons for essential public assistance benefits and services, including healthcare services, as well as to grant, reduce, revoke, or reclaim such benefits and services.
Contact

