The right balance: how to fix European Union artificial intelligence regulation
The right balance: how to fix European Union artificial intelligence regulation
Publish Date: 2026-06-11 03:18:00
Source Domain: www.bruegel.org
Here are five key points summarizing the article “The AI Act and its blind spots”:
1. The EU AI Act aims to regulate AI systems while facing challenges like the dynamic nature of AI markets, lack of enforcement experience, and global race for AI dominance.
2. The Act classifies AI systems into risk tiers and relies on principles instead of detailed obligations to keep up with fast-moving markets. Developers assess compliance themselves.
3. The Act has several blind spots: it relies on a false dichotomy between high- and low-risk systems which does not account for changing risks of AI systems after deployment. Compliance costs can be disproportionate for smaller firms. Enforcement may be fragmented and ineffective. Decisions about standards are made behind closed doors with limited input from civil society.
4. The Act lacks an ad-hoc liability mechanism which makes it difficult for victims of AI harm to obtain redress. Existing liability frameworks were not designed for AI.
5. To address these issues, the authors propose a hybrid regulatory approach with both ex-ante and ex-post elements, more tailored ex-ante requirements based on deployment scale, an effective ex-post liability system and increased ex-post transparency through third-party auditing, incident reporting and standardized harm taxonomy.
Hope these five key points help summarize the main arguments and proposals from the article! Let me know if you have any other questions.
