Using artificial intelligence in medical devices

• 1 August 2024
  • Entry into force of the AIA.
• 2 February 2025
  • Application of general provisions and prohibited AI practices.
• 1 August 2025
  • Notified bodies, GPAI models, governance, confidentiality and penalties. 
• 2 August 2026
  • The other provisions enter into force with the exception of Article 6(1).
• 2 August 2027
  • The AIA is applied in full to all AI systems, including medical devices. 

The first step is to assess whether the system meets the definition of the AI Act. According to Article 3(1) of the AIA, an AI system means a “machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments”.

The European Commission has published guidelines on the definition of an AI system. The Commission’s guidelines help operators understand whether their systems fall within the scope of the AI Act. The guidelines clarify the seven key elements of the definition of an AI system and provide practical guidance on how to apply the definition in various situations. This allows operators to assess the characteristics of their system, check the need to comply with the legislation and requirements and plan necessary measures. 

The AI Act imposes obligations on AI systems with a risk-based approach:

1. Certain AI systems and uses are prohibited.  
2. High-risk systems are subject to extensive requirements before they can be placed on the market or put into service.  
3. Certain low-risk systems are subject to transparency obligations.

There are also AI systems that are so low-risk that they are not subjected by any obligations under the AI Act. Note: general-purpose AI models are regulated separately. 

The regulation of the AI Act mainly concerns high-risk AI systems. These refer to AI systems that may have a significant adverse impact on the health, safety and fundamental rights of EU citizens. 

The European Commission has published guidelines on prohibited AI practices and the definition of an AI system. The guidelines on prohibited AI practices provide an overview of prohibited and high-risk AI systems and AI systems subject to transparency obligations and emphasise the protection of health, safety and fundamental rights. 

An operator means a provider, authorised representative, importer, distributor or deployer. 

Provider means a natural or legal person, public authority, agency or other body that develops an AI system or a general-purpose AI model or that has an AI system or a general-purpose AI model developed and places it on the market or puts the AI system into service under its own name or trademark, whether for payment or free of charge. Any reference to a “manufacturer"” in accordance with the MDR (2017/745) and IVDR (2017/746) are to be understood as references to a “provider” in accordance with the AI Act. 

Authorised representative means a natural or legal person located or established in the European Union who has received and accepted a written mandate from a provider of an AI system or a general-purpose AI model to, respectively, perform and carry out on its behalf the obligations and procedures established by this Regulation.

Importer means a natural or legal person located or established in the Union that places on the EU market an AI system that bears the name or trademark of a natural or legal person established in a third country.

Distributor means a natural or legal person in the supply chain, other than the provider or the importer, that makes an AI system available on the Union market.

Deployer means a natural or legal person, public authority, agency or other body using an AI system under its authority except where the AI system is used in the course of a personal non-professional activity.

The AI Act focuses on the obligations of AI system providers and other operators. The AIA also gives the users of AI systems certain rights. If a decision significantly affecting a person is made on the basis of a result produced by a high-risk AI system, that person has the right to obtain a clear and understandable explanation of the role of the AI system in said decision-making process. In addition, any natural or legal person who considers that the provisions of the AI Act have been breached has the right to lodge a complaint with a market surveillance authority.

AI is within the scope of the regulation on medical devices if it is an integral part of a medical device or an independent piece of software that meets the definition of a medical device in accordance with the MDR/IVDR. According to the MDCG 2019-11, medical device software (MDSW) is software that is intended to be used, alone or in combination, for a medical purpose as specified in Article 2(1) of the MDR or Article 2(2) of the IVDR. 

If a system meets the definition of an AI system in the AI Act and is an integral part of a medical device or an independent piece of software in accordance with the MDR/IVDR, both the regulation on medical devices and the AI Act apply. More information about software as medical devices is available on the relevant page on the Fimea website. This page provides additional information on when software is considered a medical device and what regulation applies to it.

A medical device artificial intelligence (MDAI) is often classified as a high-risk AI system according to the classification rules described in Article 6 of the AI Act. This means that the obligations of the AIA for high-risk AI systems must be met in addition to the requirements set by the MDR/IVDR. 

An MDAI is considered a high-risk AI-system in accordance with Article 6(1) of the AIA if it meets both of the following conditions:

1. the MDAI is a safety component or the medical device itself
2. the MDAI is subject to a third-party conformity assessment by a notified body in accordance with MDR/IVDR.

Examples of classifications involving a notified body and where the AIA conditions for high risk (Art.6(1)) are met:

• MDR Class I (sterile, measuring, reusable surgical)
• MDR Class IIa, IIb, III
• MDR Annex XVI
• IVDR Class A (sterile)
• IVDR Class B, C, D.

Examples of classifications not involving a notified body and where the AIA conditions for high risk (Art.6(1)) are not met:

• MDR Class I (non-sterile, non-measuring, not reusable surgical)
• IVDR Class A (non-sterile)
• Custom-made by a health institution according to MDR/IVDR Article 5(5).

European Union expert groups have published a guideline (MDCG 2025-6) on the interplay between the AI Act and the MDR/IVDR concerning medical devices. The guideline is formatted as questions and answers and is specifically aimed at medical device manufacturers, notified bodies and competent authorities. The guideline provides answers to questions that have arisen so far, and it is supplemented and updated as experience is gained.

Below are the requirements for high-risk AI systems and their links to MDR/IVDR requirements from the perspective of a provider’s obligations. Note: any reference to a “manufacturer” in accordance with the MDR/IVDR are to be understood as references to a “provider” in accordance with the AI Act. All requirements for high-risk AI systems, as well as the obligations of providers and deployers of high-risk AI systems and other parties, are in Chapter 3, Sections 2 and 3 of the AI Act.   

Management systems:

The AI Act introduces additional requirements for the management systems of high-risk MDAI in addition to the requirements of the MDR and IVDR. These include a system-specific quality management system (AIA, Article 17), which includes data management, transparency and human oversight, and continuous and iterative risk management (AIA, Article 9), which also covers fundamental rights risks and data biases. In addition, the AI Act places particular emphasis on secondary market monitoring for systems that learn after being put into service.

Technical documentation: 

The MDR, IVDR and the AI Act require the provision of comprehensive technical documentation on MDAI systems. The MD and IVD regulations require detailed descriptions of software, software architecture, data processing methods and risk management strategies. The AI Act, on the other hand, requires additional documentation, focusing on transparency and responsibility. This includes risk assessments, information management practices and performance test results of high-risk MDAI systems. The AI Act requires preparing a single set of technical documentation for high-risk MDAI systems, containing the information required in both the medical device regulations and the technical documentation of the AI Act (AIA, Article 11, paragraph 2). The MDCG/AIB Guideline 2025-6 provides practical advice on how to combine the requirements of the MDR/IVDR and the AI Act into a single technical file.

Data and data management:

The MDR and IVDR require that clinical data be reliable and represent the intended use of the device. The AI Act, on the other hand, sets requirements for the quality, management and transparency of data (AIA, Article 10). The training, validation and test data of high-risk MDAI systems must have appropriate statistical characteristics and be as accurate, representative and appropriate as possible. Data protection and the original purpose of the data must be documented transparently. Manufacturers must implement procedures that ensure data integrity and the identification and prevention/mitigation of potential biases.

Collecting logs:

The AI Act requires for a high-risk MDAI system to automatically store log data throughout its lifecycle to ensure traceability (AIA, Article 12). This ensures that the functioning of the system is monitored, that risks affecting fundamental rights are monitored and that post-market surveillance is carried out.

Transparency and human oversight:

The AI Act and the MDR/IVDR impose obligations in terms of transparency and human oversight. Manufacturers must design and develop high-risk MDAI systems with sufficient transparency to allow users to correctly interpret the output and use the system properly (AIA, Article 13). This requires clear and comprehensible user instructions, information on the features, limitations and operating logic of the system, and documentation that supports the explainability of decisions. In addition, the AI Act calls for technical solutions that enable human intervention in critical decisions and ensure that the system cannot override these restrictions (AIA, Article 14). These obligations are complementary to the basic requirements of the MDR/IVDR, which highlight clear and comprehensive information on the purpose, function and limitations of a device, as well as documentation supporting traceability and safe use.

Precision, stability and cybersecurity:

The MDR, IVDR and AI Act emphasise the importance of cybersecurity. Manufacturers must implement technical solutions to prevent unauthorised access, attacks and manipulation. The AIA requires high-risk MDAI systems to include technical solutions to AI-specific vulnerabilities (AIA, Article 15). Cyber security is part of risk management, quality management and conformity assessment. The manufacturer of an MDAI system must secure the transfer and storage of data, prevent the manipulation of models and respond to cybersecurity incidents. These requirements apply to the entire life cycle and operating environment of a device.

Key responsibilities of the providers of high-risk AI systems and other operators

All the responsibilities of operators with regard to high-risk AI systems are stated in Chapter 3, Section 3 of the AI Act.   

Main obligations of a provider

• implementation of risk and quality management systems
• record-keeping and compliance with transparency obligations
• system testing and conformity assessment before placing on the market or putting into service
• data quality and technical documentation
• CE marking and EU declaration of conformity
• registration obligation
• traceability and human oversight
• ensuring accuracy, stability and cybersecurity
• implementing corrective measures
• cooperation with competent authorities.

Main obligations of authorised representatives

• performing the tasks specified in the mandate
• keeping the necessary documents and ensuring that they are available to the competent authorities
• registration obligation as necessary
• cooperation with competent authorities.

The mandate of an authorised representative may not include tasks such as the design and manufacturing of an AI system, the preparation of technical documentation or conformity assessment.

Main obligations of an importer

• ensuring that the system meets regulatory requirements before being placed on the market
• keeping the necessary documents and ensuring that they are available to the competent authorities
• taking corrective measures 
• informing the competent authorities if non-compliance is detected
• cooperation with competent authorities.

Main obligations of a distributor

• ensuring that the system meets regulatory requirements before being placed on the market
• keeping the necessary documents and ensuring that they are available to the competent authorities
• taking corrective measures 
• informing the competent authorities if non-compliance is detected
• cooperation with competent authorities.

Main obligations of a deployer

• using the system in accordance with its user instructions and purpose
• monitoring the operation of the system and reporting irregularities.