AI in healthcare: Catalonia's latest guidelines

On March 19, 2025, the Health/AI Program (Programa Salut/IA) –established by Resolution on March 19, 2023, and integrated into the Catalan Health System– released four new guidelines on the use of artificial intelligence tools in the healthcare sector. These latest guidelines complement the existing ones published by the Health/AI Program, including the Guide for the evaluation of digital health technologies incorporating artificial intelligence (AI), which was published on November 25, 2024, and the Guide of best practices for developing code in artificial intelligence health solutions, released in January 2023.

While these documents are directed at all stakeholders involved in the development and implementation of AI algorithms within the Catalan Health System, the guidelines are based on recent European regulations, making them relevant and valuable to Spanish and European stakeholders alike.

Their main purpose –according to the Generalitat de Catalunya’s website–, is to support technological innovation in the healthcare sector, ensuring that AI-based technologies are developed safely, reliably, and in accordance with current European and local regulations. To this end, they offer practical guidance for professionals and organizations within the sector, along with explanatory infographics to enhance comprehension.

For more detailed information and practical guidance on the AI Act and its implications and effects, we recommend considering our recent blog posts, which, although not specifically referring to the healthcare sector, provide a very useful guide to understanding the scope and application of the AI Act, the AI practices prohibited by the regulations and the legal definition of AI system.

Below, we highlight some of the key points included in each guideline:

• Guide for good practices in generative AI in the healthcare sector: Provides clear guidelines for the development and implementation of generative AI tools, specifically Large Language Models (LLM), in the healthcare sector. The guide emphasizes the importance of ethical and responsible use of AI to improve health outcomes while ensuring compliance with legal and regulatory standards. It underscores the necessity of human supervision, data quality, interpretability of results, and privacy protection in the deployment of these technologies.
  
  
• Guide for the application of the AI Act in the healthcare sector: Addresses the interaction of the AI Act with other regulations, such as the Regulation (EU) 2017/745 of the European Parliament and of the Council of 5 April 2017 on medical devices (the Medical Device Regulation), Regulation (EU) 2017/746 of the European Parliament and of the Council of 5 April 2017 on in vitro diagnostic medical devices (the In Vitro Diagnostic Regulation), and the Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data (the General Data Protection Regulation).
  
  The document concludes with a step-by-step process for bringing new AI systems to market or maintaining existing systems, including conformity assessment, CE marking, and registration in relevant databases. It also provides detailed annexes outlining specific obligations for different stakeholders involved in high-risk AI systems. 
  
  
• Guide for AI software qualification as a medical product: Outlines the process for determining whether AI software qualifies as a medical device under the MDR and IVDR. It details the regulatory context, risk classification, and compliance requirements necessary for AI software to be considered a medical product.
  
  The guide emphasizes the importance of accurately defining the intended purpose of the software, as this impacts its qualification and classification. The document provides a systematic approach to qualifying and classifying medical device software (MDSW), including detailed steps to determine if the software falls under the MDR or IVDR. It also addresses the application of risk classification rules (whether the medical device is from class I, Is, Im, IIa, IIb or III or the IVDR product from class A, As, B,C or D) and the importance of ensuring compliance with safety and performance standards.
  
  
• Guide for obtaining a CE marking for AI medical software: Serves as a practical guide for obtaining CE marking for AI-based medical software. It details the regulatory requirements, classification, technical documentation preparation, quality system implementation, and conformity assessment processes in accordance with the MDR and IVDR.
  
  The guide also covers the registration of products and operators in the EUDAMED database and the Agencia Española de Medicamentos y Productos Sanitarios. The document emphasizes the importance of maintaining compliance with regulatory standards throughout the software's lifecycle, including post-market surveillance and periodic safety updates.

With the introduction of these new guidelines, the Health/AI Program has set a robust framework for the responsible use of AI in healthcare. These guidelines not only align with European regulations but also provide practical steps for developers and manufacturers to navigate the complex regulatory landscape. By focusing on ethical considerations, data quality, and patient safety, the guidelines aim to foster trust and reliability in AI-driven healthcare solutions, paving the way for future advancements in the field.