EU AI Act - Welche KI-Regeln jetzt gelten - und welche (noch) nicht
Summary
Hosts Karina Schröder and Ralf Kraunder discuss the EU AI Act’s delayed implementation, focusing on the postponement of high-risk AI regulations by 16 months due to the "Digital Omnibus" package. They analyze how tech lobbying influenced these changes, weakening oversight while maintaining bans on non-consensual intimate images. The episode highlights that transparency rules for AI-generated content still take effect in August 2026, serving as a critical test for EU regulatory ambition.
Topics discussed
Intro: Hypothetical book delay and the EU AI Act postponement
Welcome and overview of the EU AI Act's status and delays
Banned AI applications and rules for foundation models
High-risk AI systems and the Belgian welfare scandal
The Digital Omnibus package and reasons for the delay
Weakened training data rules and industry lobbying
New bans on manipulative AI and deepfake abuse
Watermarking and labeling requirements for AI content
Lobbying influence and the balance between regulation and innovation
Case study: AI in social welfare and final thoughts
Outro and teaser for the next episode
Listen ad-free on Castria