Implementation strategies for the AI Act

1. Conduct a comprehensive gap analysis

The first step is to identify all AI components in your products and classify them according to the risk-based framework of the AI Act. In addition, are your existing technical documentation, risk management processes and market surveillance systems sufficient to meet the requirements of the AI Act? This will allow you to identify areas for improvement, such as expanding risk management to include AI-specific risks (e.g. bias).

2. Develop a standardized compliance framework

Manufacturers should integrate the requirements of the AI Act and the MDR into a single framework to avoid duplication of effort and to facilitate the compliance process. Harmonized technical documentation covering both regulations is particularly important. For example, the technical documentation for an AI-assisted medical device should now cover not only the clinical safety and performance of the device, but also information on the functionality of the AI system, the quality of the training data, validation processes and risk management strategies for AI-specific hazards such as bias or faulty algorithms.

3. Employee training

Compliance with AI regulations also requires a cultural and procedural change within organizations. Manufacturers must ensure that all relevant employees, especially those in regulatory affairs, software development and risk management, are thoroughly trained in the requirements of the AI Act and AI. Employees should be familiar with expectations around transparency, risk management, data management and training. A well-trained team will ensure that the regulations are embedded in their organization, reducing the risk of errors in regulatory assessment.

4. Early collaboration with notified bodies

Notified Bodies also face this challenge, as they too must first enable their staff to fully assess AI systems. In addition, Notified Bodies will need to be redesignated by the Certification Authority under Article 31 in order to be authorized to approve AI products at all. Manufacturers should first clarify with their Notified Bodies whether they are able to assess the requirements or whether they need to look for another body.

Manufacturers should then communicate with their Notified Bodies early in the development phase to clarify the criteria for AI-based systems. Early collaboration will ensure that the notified body understands the nuances of the AI component within their medical device, including how the AI will be trained and validated, and how it will be continuously monitored once implemented. This helps get both parties on the same page and avoids delays and potential rework.

5. Get involved in the development of the guidelines

Use your networks, get involved and play a role in improving the guidelines. As a medical device manufacturer, we can provide legislators with industry insights and support practical measures. In this way, the important and necessary additional safety precautions of the AI law can be implemented without the additional effort causing useful products to disappear from the market or slowing down innovation, as has often been said about the MDR regulations.

A good approach is the AI PACT, in which companies have come together to promote dialogue between all stakeholders and, above all, to work proactively on the implementation of the AI Act. This not only has the advantage of active participation, but also automatically implements all the above points for the participating companies.

These recommendations can help you better prepare for the combined requirements of the MDR and the AI Act, and bring your future AI-powered devices to market safely and without regulatory gaps. In the third part of our ‘AI Act Insights’ series, we will take a closer look at the specific responsibilities of manufacturers developing AI-powered medical devices and how you can ensure your products are ready for the market.