Artificial Intelligence in Medical Devices: new regulatory challenges between MDR and AI Act

Artificial intelligence (AI) is playing an increasingly central role in the development of medical devices, with applications already well established in supporting healthcare professionals and improving diagnostic accuracy. Alongside these opportunities, however, new regulatory complexities are emerging, requiring a structured approach to compliance.

On this topic Gian Luca Brusati, Regulatory Affairs Consultant at ISEMED, contributed an in-depth article published in T-Magazine, focusing on the impact of the new Regulation (EU) 2024/1689 – Artificial Intelligence Act (AIA) on AI-based medical devices.

Read the last issue of T-Magazine: HERE

 

The regulatory framework: MDR and AI Act compared

Until now, the main regulatory reference for medical devices has been Regulation (EU) 2017/745 (MDR), which governs the design, manufacturing, and post-market surveillance of medical devices placed on the European market.

With the publication of the Artificial Intelligence Act (AIA) on July 12, 2024, the regulatory landscape has expanded to include specific requirements for artificial intelligence systems. In particular, medical devices that:

• use AI algorithms
• require the involvement of a Notified Body

are classified as high-risk systems and must comply with both the MDR and the AIA.

The AIA will become applicable to this category from August 2, 2027.

High-risk medical devices: what changes for manufacturers

The AI Act introduces additional obligations compared to the MDR, raising the level of responsibility for manufacturers. The most relevant aspects include:

Data governance
The data used in the design and validation of AI-based medical devices must be:

• representative
• complete
• free from bias

Patient diversity (gender, ethnicity, socio-economic conditions) must be taken into account to ensure safety and effectiveness. Data governance procedures must be properly integrated and rigorously documented.

Transparency

Manufacturers are required to provide:

• clear and understandable instructions on device operation
• information on limitations and conditions of use

This enables users to make informed decisions and correctly interpret the outputs generated by AI systems.

Accountability and traceability
Medical devices must ensure:

• traceability of events
• systematic recording of operational data

These requirements support process reconstruction, clear attribution of responsibility, and timely intervention in the event of malfunctions or incidents, in line with MDR post-market surveillance provisions.

Technical documentation and quality management system

A key aspect highlighted in the article concerns documentation integration. The AI Act allows manufacturers to:

• integrate AIA technical documentation (Annex IV) with existing MDR documentation
• avoid duplication of overlapping sections

For SMEs and start-ups, the Regulation also provides for the possibility of submitting simplified documentation, based on a template that is currently under development.

With regard to the quality management system, Article 17 of the AIA requires implementation to be proportionate to the size of the organization.
While ISO 13485 already covers a substantial part of the general QMS requirements, additional AI-specific elements must be integrated, including:

• processes for data collection, cleaning, labelling, and input into AI models
• management and retention of operational records
• management of AI-specific risks
• additional considerations for post-market surveillance activities

ISEMED’s support in the compliance journey

The ISEMED regulatory team has extensive experience in supporting medical device manufacturers with MDR compliance. This expertise now extends to the Artificial Intelligence Act, enabling companies to navigate both regulatory frameworks through a:

• integrated
• pragmatic
• future-oriented approach

This structured methodology supports manufacturers in managing the transition to the new European regulatory environment.

Preparing today for the AI Act’s entry into force

While harmonized standards and implementation guidelines are still forthcoming, manufacturers are encouraged to begin preparing now by:

• planning compliance activities
• allocating dedicated resources
• updating internal processes and system procedures

Achieving dual compliance with MDR and AIA represents a complex challenge, but also an opportunity to strengthen the quality, safety, and reliability of AI-based medical devices.