Regulation of the use of AI by industry

AI used for a medical purpose commonly falls within the definition of a medical device. The MHRA is in the process of updating the UK's existing framework for medical device regulation (the Medical Devices Regulation 2002) as we have discussed here and, whilst the existing system of risk-based classification of medical devices will continue in the updated regime, many AI products will be up-classified in order to further protect users and patients.

Alongside the anticipated reform of the UK medical devices regulation, the MHRA is progressing a number of other programmes focused on AI as a medical device and/or software as a medical device(AIaMD/SaMD):

• the imminent launch of the AI-Airlock, a regulatory sandbox for AIaMD; and
• the publication of a series of guidance notes on AIaMD/SaMD, with two further publications on "Good machine learning practice for medical device development" and "Best practice AIaMD development and deployment" in the pipeline (with no set timeline) and guidance on cybersecurity due by Spring 2025.

However, the MHRA publication also states that, in many cases, the adoption of AI by industry will not impact how the MHRA regulates - "the questions that [the MHRA] needs to ask to determine whether a product is safe do not change when the nature of the evidence we consider changes"'.  

Some uses of AI by industry will not fall within the definition of a medical device, for instance the use of AI in drug discovery or as part of a clinical trial protocol. Nonetheless, as AI becomes more widespread throughout medical product lifecycles, both the MHRA and industry need to be able to assess and act on results driven by AI within the parameters set by regulation.

Use of AI by the MHRA

The MHRA publication sets out the steps that the MHRA is taking to make use of the opportunities presented by AI to enhance its performance as a public service organisation delivering time-critical decisions which impact public and patient safety.

The MHRA is in the nascent stages of implementing AI within its own processes and is doing so in line with existing government strategies (including the National AI Strategy and NHS National Strategy for AI in Health and Social Care). The publication highlights areas where the MHRA is exploring whether AI could benefit existing processes, including:

• the use of supervised machine learning to conduct an initial assessment of materials submitted as part of a marketing authorisation application, to reduce the need for human involvement in the early assessment stages;
• exploring the role that generative AI and large language models might play in medical regulation and approaches to obtaining/analysing real-world data; and
• screening for fraudulent medicinal products and medical devices sold online.

Conclusions

The key takeaways from the publication include that AI touches all aspects of the MHRA’s remit, from regulating medical products to how it operates and make decisions as a regulator.  The MHRA’s intention to follow the pro-innovation approach to AI set out in the AI White Paper while also ensuring patient safety is clear throughout. The MHRA is also committed to taking an internationally aligned approach to regulating AI and developing international best practice, with the UK co-chairing the International Medical Devices Regulators Forum (IMDRF) working group on AI and Machine Learning enable medical devices.  

The upcoming changes to the UK Medical Devices Regulations, the various guidance documents in development and the MHRA’s ongoing AIaMD/SaMD programme all provide opportunities for the MHRA’s strategic approach to AI to be implemented in practice.