The European Union is making history in the regulation of artificial intelligence: it is the first region in the world to develop concrete practical guidelines for the development and use of general-purpose AI - versatile AI models that are behind systems such as ChatGPT. Matthias Samwald, Associate Professor at the Institute for Artificial Intelligence at the Medical University of Vienna, is co-chairing one of the four central working groups.
"We are at a crucial point in the development of artificial intelligence. In the biomedical field, we see the enormous potential of general-purpose AI models, for example in accelerating research or improving medical diagnoses. At the same time, however, we must also take the systemic risks seriously - from the danger of large-scale disinformation to fundamental challenges in the control of autonomous AI systems," says Samwald, explaining his motivation for the commitment.
The "General-Purpose AI Code of Practice" will translate the fundamental principles of the EU AI Act into practical, implementable measures. In his role as co-chair of the "Risk identification and assessment" working group, Samwald and his team are defining criteria for identifying and assessing AI models that could pose systemic risks.
"The EU is currently being criticized for over-regulation, but a practicable set of rules can create benefits for innovation and safety," Samwald emphasizes. "Unlike in the US, for example, where different states develop different rules, we want to provide companies and with a clear framework that applies across the EU. This may not only reduce the costs of developing AI systems, but also enable more effective risk prevention."
The Code of Practice is being developed in a unique consultation process: around 1000 stakeholders from business, science and civil society are involved. The first draft version was published on November 14, 2024. The final version should be available by May 2025.
About the person
Matthias Samwald is an Associate Professor at the Institute of Artificial Intelligence at the Medical University of Vienna. His research focuses on the use of AI to accelerate biomedical research and improve healthcare. He was co-initiator of the ¤15 million "U-PGx" project on personalized medicine and contributes his expertise in areas such as large language models, system evaluation and AI governance.
