The EU regulatory landscape for medical devices making use of artificial intelligence (MDAI) has become significantly more complex with the introduction of the EU AI Act. The Medical Device Coordination Group (MDCG) and the Joint Artificial Intelligence Board (AIB) have now published comprehensive guidance (AIB 2025-1/MDCG 2025-6) entitled “Interplay between the Medical Devices Regulation (MDR) & In Vitro Diagnostic Medical Devices Regulation (IVDR) and the Artificial Intelligence Act (AIA)” (the guidance) in June 2025. This guidance aims to provide clarity for manufacturers navigating the dual regulatory framework that now governs AI-enabled medical devices.

This article provides a brief summary of the guidance, practical next steps, and how these may apply to manufacturers considering entry into the Great Britain market.

1. European Union

Dual Regulatory Framework

The MDR and IVDR requirements address risks related to medical device software; however, they do not explicitly address risks specific to AI systems. The AIA complements the MDR/IVDR by introducing requirements to address hazards and risks to health, safety, and fundamental rights specific to AI systems.

This dual application means manufacturers must ensure compliance with both frameworks simultaneously. Manufacturers of MDAI may choose to integrate, as appropriate, the necessary testing and reporting processes, as well as the information and documentation required for their MDAI, into the documentation and procedures already established under the MDR/IVDR.

In practice, alongside the MDR/IVDR and the AIA, numerous other EU regulations will often apply to these devices. Taking this into consideration, streamlining the different processes is of paramount importance.

Quality Management Systems

The AIA mandates that providers and manufacturers implement quality management systems (QMS) for AI systems to ensure compliance with the AIA, covering substantive requirements and procedural obligations, implemented in a proportionate manner according to the size of the provider’s organisation. Manufacturers of AI systems may incorporate the elements of the QMS required by the AIA within their existing QMS established under the MDR/IVDR.

This integration approach extends to risk management systems, where manufacturers of high-risk MDAI may integrate the additional risk management requirements specific to MDAI, as set out in paragraphs 1–9 of Article 9 of the AIA, into their existing documentation and procedures under the MDR/IVDR.

Data Governance and Bias Mitigation

The AIA introduces specific requirements for data governance that complement existing MDR/IVDR clinical data requirements. The AIA requires high-risk MDAI manufacturers to implement data governance and management practices appropriate for the intended purpose, including addressing possible biases likely to affect health and safety, have a negative impact on fundamental rights, or lead to discrimination prohibited under Union law. The MDR requires, more generally, that the clinical data relied upon is sound and robust.

While the AIA provides additional safeguards, the guidance notes that the aim of both pieces of legislation is largely aligned in this respect.

Transparency and Human Oversight Requirements

Under the AIA, transparency is a core requirement for high-risk MDAI. It imposes a legally binding obligation on providers to design and develop high-risk MDAI in such a way that their operation is sufficiently transparent to enable deployers to interpret outputs correctly and use the system appropriately. Under the MDR, several related requirements are included within the General Safety and Performance Requirements and Annexes I–III.

The AIA emphasises, and legally mandates, that manufacturers design and develop AI systems, including high-risk MDAI, with appropriate human oversight mechanisms. The MDR requires that devices are designed for safe and effective use.

In summary, the guidance concludes that both pieces of legislation are broadly complementary in this regard, reinforcing the importance of these requirements.

Conformity Assessments

For AI systems classified as high-risk MDAI under Article 6(1) (i.e. those requiring third-party conformity assessment), the relevant conformity assessment procedure is determined by the MDR/IVDR. For high-risk MDAI where both I and III of the AIA apply, Annex I of the AIA prevails.

High-risk MDAI undergo the relevant conformity assessment procedure based on the device’s risk classification under the MDR and IVDR. Most MDAI are classified as Class IIa (MDR), B (IVDR), or above, meaning that a notified body must conduct the assessment.

If an MDAI is classified as high-risk under the AIA pursuant to Article 6(2), the AIA prescribes the applicable conformity assessment procedures.

Substantial Modifications and Change Control

The AI Act introduces the concept of a “substantial modification”, which manufacturers must carefully consider. This is defined autonomously under Article 3(23) of the AIA. In accordance with Article 43(4), high-risk MDAI that have undergone a conformity assessment must undergo a new conformity assessment in the event of a substantial modification.

However, changes to high-risk MDAI that have been pre-determined by the manufacturer and assessed during the initial conformity assessment - where included in the technical documentation referred to in AIA Annex IV point 2(f) - will not constitute a substantial modification.

Although not explicitly mentioned in the guidance, it may prove challenging to align the AIA framework for substantial modifications with the existing MDR framework for substantial changes.

Timeline for Compliance

The date of application of obligations for “Annex I high-risk AI systems” is 2 August 2027. This means that if a medical device has been placed on the market or put into service before 2 August 2027 and undergoes any significant design change on or after that date, the EU AIA obligations will apply.

2. Great Britain

The UK Framework

The AIA does not apply in the UK. The UK regulations governing medical devices placed on the Great Britain market, including MDAI, are under review as part of the wider programme of reform to medical device regulation. Significant changes to this landscape are expected in the near future.

While there has been extensive discussion regarding MDAI, no binding updates have yet been made to existing regulations to address the complexities of AI in medical devices. AI-specific guidance and commentary have, however, been published on several related topics.

Currently, the applicable regulatory framework and guidance for MDAI are the same as for Software as a Medical Device (SaMD). Accordingly, the overarching framework in Great Britain (England, Wales, and Scotland) is the UK Medical Devices Regulations 2002 (as amended) (UK MDR). The primary guidance applicable to SaMD, as published by the MHRA, also applies to MDAI.

According to the MHRA’s policy paper of 30 April 2024, where AI is used for a “medical purpose”, it is very likely to fall within the definition of a general medical device. This means that it must meet the requirements of the UK MDR before being placed on the Great Britain market.

In short, the MHRA proposes to deliver regulatory reform in this area through updated regulations and guidance, more streamlined processes for SaMD/MDAI, and the potential up-classification of such devices (from the current Class I). This area remains in flux and should be monitored closely.

UK MDAI Initiatives

What follows in this section is a snapshot of some initiatives that may impact upon the UK MDAI regulatory landscape, including any practical AI-specific considerations that MDAI developers should bear in mind. 

In the MHRA’s AI White Paper, a number of key principles are addressed from an AI perspective, including:

• Transparency and explainability of AI: All devices, including MDAI, must be designed with the intended users in mind, and manufacturers must provide a clear statement of the device’s purpose. A key risk for MDAI is the human–device interface. Existing MHRA guidance on applying human factors to medical devices is being supplemented by further detailed guidance specific to MDAI.
• Fairness: The MHRA is committed to ensuring equitable access to safe, effective, and high-quality medical devices for all users. It intends to take an internationally aligned position, encouraging manufacturers to refer to ISO/IEC TR 24027:2021 (Information technology – Artificial intelligence – Bias in AI systems and AI-aided decision making) and IMDRF guidance document N65.
• Accountability and governance: These principles apply to the datasets used to create AI models and any changes made post-market. In collaboration with the US Food and Drug Administration (FDA) and Health Canada, the MHRA has published guidance on Predetermined Change Control Plans (PCCPs) to enable full traceability and accountability for how AI models meet intended use and the impact of changes.

The MHRA has also (re)launched a regulatory sandbox for MDAI, known as the AI Airlock. Using real-world products, the AI Airlock brings together expertise from within the MHRA and partners such as UK Approved Bodies, the NHS, and other regulators. The outputs of this initiative will inform future MHRA guidance and policy, while exploring limitations of current approaches to demonstrating regulatory compliance for MDAI. The second phase of the AI Airlock is currently ongoing.

Looking Forward

The MHRA is developing supplementary guidance to ensure AIaMD placed on the market is supported by robust assurance regarding safety and effectiveness, and outlining technical methods to test MDAI. Key topics for manufacturers include:

Good Machine Learning Practice (GMLP): Based on 10 guiding principles jointly developed by the MHRA, FDA, and Health Canada, intended to form the foundation for GMLP addressing the unique nature of MDAI. Future MHRA guidance will map these principles to key responsibilities under the UK MDR.

Predetermined Change Control Plans (PCCPs): Addressing how manufacturers can manage post-market AI updates, this guidance aligns internationally with the five guiding principles developed with the FDA and Health Canada.

Transparency of machine learning medical devices: The MHRA, FDA, and Health Canada have jointly established guiding principles on transparency, offering good practice guidance applicable to all medical devices.

3. Next Steps and Practical Recommendations

For manufacturers, the combination of the AIA and MDR requires forward planning -something many will already have initiated. In light of the guidance, the following actions should be prioritised to understand where your medical device fits within this evolving framework:

Conduct an AI Act risk assessment: Determine whether your MDAI products qualify as high-risk AI systems based on their MDR/IVDR classification and notified body involvement.

Review your QMS: Assess opportunities to integrate AI Act requirements into existing MDR/IVDR quality management systems to avoid duplication.

Evaluate data governance practices: Ensure frameworks address AI-specific bias detection and mitigation requirements.

Plan for substantial modifications: Develop change control procedures that account for the AI Act’s “substantial modification” requirements while leveraging pre-determined change provisions.

Given the contrasting regulatory frameworks currently applicable to MDAI in the EU and UK, manufacturers should carefully consider their regulatory compliance strategy, which will depend on, among other factors, their target European markets. With the UK Government’s proposed plan to recognise CE-marked medical devices indefinitely (pursuant to conformity assessment under relevant EU regulations) for the Great Britain market - alongside other potential international recognition mechanisms - manufacturers may be able to streamline their MDAI compliance efforts across both the EU and UK.