EU AI Act: Key Takeaways

The EU's AI Act, effective Aug. 1, regulates AI through phased implementation and compliance obligations. Enforcement will be robust, with national authorities and the EAIB overseeing compliance.

Non-compliance can result in fines up to 35M euros or 7% of global revenue. The Act complements GDPR, prohibiting AI that exploits vulnerabilities or creates non-consensual facial recognition databases.

It also imposes new requirements for general-purpose AI by August 2025 and rules for high-risk AI systems by August 2026. Each EU member state must appoint a national authority to oversee compliance. The debate in Germany centers around which agency will take on this role.

Privacy advocates suggest that the data protection authorities are well-suited for the task. However, experts argue for a centralized approach, with the Bundesnetzagentur being a potential candidate.

Concerns about diffuse responsibilities and a lack of clarity in the regulatory landscape persist. The German Bundestag has limited time to address oversight of the AI Act, and establishing a new authority may prove cumbersome.

The implementation and enforcement of the AI Act will continue to be a topic of discussion and refinement.