Artificial Intelligence – Data inputs must be the focus of AI and its regulation

The European Union is preparing to pass the world's first artificial intelligence law, the AI Act. While this regulation focuses on the risks associated with AI use, researchers suggest that it would be more effective to focus on the quality of input data to ensure compliance and safety.

The EU's forthcoming AI Act, due in April, will establish safeguards for AI tools, including generative models such as ChatGPT or Gemini. By assessing the risks of using AI, this law will require the use of high-quality input data. However, the implementation of this requirement remains unclear, requiring an alternative method to ensure compliance.

Christian Peukert, Associate Professor at HEC Lausanne, proposes focusing on input data to reduce compliance risks. His team recommends apportioning responsibility for poor-quality data sources between AI developers and deployers, based on exogenous or endogenous data issues. This approach could improve data quality, reduce AI bias and provide appropriate incentives for effective AI regulation.

Learn more