EU AI Act: Compact practical guide for the life sciences sector

From research and diagnosis to the development of innovative therapies: Artificial intelligence (AI) has the potential to transform various segments of the life sciences industry. Thanks to technological progress, it will be possible in future to analyse large volumes of biomedical data, recognise complex patterns and make precise predictions. This is a significant value lever that opens up new growth options.

The European Union's AI Act (EU AI Act), the future legal framework for the safe and ethical use of AI, is legally groundbreaking - and binding - for companies. But what exactly does the EU AI Act mean for the life sciences sector? In a white paper, our experts now show precisely and practically which requirements and implications companies should keep a close eye on when preparing for compliance - and how a proactive approach to risk minimisation can be designed. The title of the publication: "A compass for the Artificial Intelligence Act (AI Act)".

The EU AI Act categorises AI systems according to risk level and imposes obligations on developers and users: For AI systems with a high risk as defined by the law, developers must undergo a so-called conformity assessment before implementation. This also applies to systems that are used in medical diagnosis, treatment recommendations and precision medicine, for example. Although self-assessment is envisaged in principle due to the lack of harmonised standards, this will only be possible in practice in the future.

The EU AI Act also applies to AI systems that are developed outside the EU but marketed or used in the EU. This means that developers of AI systems must comply with the EU AI Act regardless of their location if they wish to offer their systems in the EU. In addition, the EU AI Act is also interpreted to apply to life sciences companies from third countries that use artificial intelligence to develop products to be sold on the European market.

The EU AI Act is expected to come into force in June 2024. As it currently stands, the Act provides for graduated transition periods of between six and 36 months for different categories of AI systems. Managers working in the areas of compliance, data governance and the development and use of AI technologies should be prepared for new responsibilities.

The EU AI Act has far-reaching consequences for various areas of the life sciences sector, including companion diagnostics and clinical trials. The EU AI Act will also have a significant impact on the medical devices industry, particularly on the manufacture of products that are categorised as "high-risk". The requirements relate in particular to risk management, data governance and compliance monitoring. For companies that manufacture AI-supported products, it is advisable to seek external advice for the holistic management of the tasks in view of the complex situation.

In the white paper, our experts break down the nine categories that need to be assessed in particular in view of the future requirements of companies with regard to their level of maturity. The individual aspects are presented separately. The assessment can then be made on a scale from 1 (unprepared) to 5 (leading). The following aspects must be categorised:

• Risk classification
• Conformity assessment
• Transparency
• Human oversight
• Fairness, non-discrimination and bias
• Explainability
• Data governance
• Cyber security
• Compliance and enforcement

Our authors then go on to explain in detail how KPMG can support companies in the life sciences sector - from strategy development to process monitoring in practice.