Amendments to the EU AI Act
A policy memo with the scope of improving the current version of the EU AI Act in a social sustainability framework.
Challenge
Does the EU AI Act promote socially sustainable AI?
The challenge involves a detailed analysis of current version of the EU AI Act with a focus on the question: In how far does the EU AI Act promote social sustainability goals? The goal is to provide some suggestions for improvement to be forwarded to MEP and shadow rapporteur, Alex Voss.
Team
Silvia Cirolini, Özge Demir, Fatma Ben Yahia, Marco Mata, Lisa Putter
Contact: silvia.cirolini@tum.de
About the prototype
The outcome of this challenge was a policy memo with more than ten amendments that have the goal of improving and ensuring the use of social sustainable AI system in the EU. The policy recommendations cover different areas of the EU Regulation such as the extension of prohibited AI systems to facial and emotion recognition systems used in border control management (Article 5); the obligation for sandbox environments to minimize the risks to discriminatory outcomes related to potential unconscious biases, health and safety (Articles 53&54), the inclusion of well known social, environmental, and economic organizations as authorities within the AI Office: for example the European Environment Agency – EEA (Article 57) and the introduction of a shutdown mechanism in case the high-risk AI systems exhibit erroneous behavior or produce incorrect answers with significant implications
Outputs
Further information about the progress, milestones, and roadblocks.
Extract from the EU Amendment
Extract from the EU Amendment
Extract from the EU Amendment