EU delays most AI Act enforcement rules to 2027–2028 while keeping labeling requirements on track for August 2026

The European Commission, European Parliament, and Council of the EU agreed on amendments bundled in the 'Digital Omnibus on AI,' pushing rules for AI in biometrics, critical infrastructure, education, and migration to December 2027, and rules for AI in products like lifts or toys to August 2028. AI systems that generate sexually explicit content without consent, including nudification apps, are now explicitly banned.

Labeling obligations under Article 50 remain largely unchanged and take effect August 2, 2026, requiring companies to label deepfakes and certain AI-generated text. The text labeling rule applies only to fully automated content that no human has reviewed or edited.

Small and medium-sized enterprises with up to 750 employees and 150 million euros in revenue receive relief, including reduced registration and documentation requirements and better access to regulatory sandboxes (test environments where companies can try out AI under real-world conditions).