The European Union's AI Act was supposed to be the world's template for governing artificial intelligence. The high-risk provisions — covering AI in healthcare, law enforcement, hiring, and critical infrastructure — were scheduled to take effect in August 2026. MEPs have now voted to support pushing that deadline to as late as December 2027. [1]
The postponement reflects a political reality the Commission did not anticipate when it passed the Act in 2024: the AI industry moved faster than the regulatory apparatus. Industrial AI applications — factory automation, supply chain optimization, predictive maintenance — may be moved outside the high-risk scope entirely, a concession to European manufacturers warning that compliance costs would drive development to the United States and China. [2]
Not everything was delayed. MEPs approved a ban on AI "nudifier" systems — tools that generate non-consensual synthetic intimate imagery. That provision takes effect immediately.
Meanwhile, the regulatory landscape is fracturing along national lines. In the United States, 45 state legislatures have introduced approximately 1,500 AI-related bills, creating a patchwork that no single company can navigate coherently.
The EU wanted a single global standard. What it got is a postponement at home and fragmentation abroad. The regulation is coming. Just not yet.
-- Anna Weber, Berlin