Blogged: AI hot mess – German courts on Meta’s AI training and LLM regulation in the EU
Paulina Jo Pesch has published a long blog post (CR-online blog) on the German court procedures regarding Meta’s AI training on social media data. The article provides
- a detailed overview of the timeline of events with links to all relevant public documents,
- new unsettling insights into the procedure in Cologne,
- and an analysis of the most important aspects of the judgments from the Higher Regional Courts of Cologne and Schleswig.
The procedures raise serious worries about the implementation of EU regulatory law in the realm of AI, and points out that courts and regulators must hold AI companies accountable. It must be doubted that AI company’s attempt to create superintelligence through training LLMs on vast amounts of often sensitive personal data of often dubious quality will be successful. However, as privacy researchers still have a very limited understanding of risks and the efficacy of mitigation measures, the stakes are high.