EU Moves AI Act Into Enforcement Phase With New Transparency and Inspection Powers

Overview

• On Sunday, August 2, 2026, the EU activated a new phase of the AI Act that brings Article 50 transparency rules into force and starts active supervision of AI providers.
• Providers of chatbots and other interactive systems must now inform users when they are talking to an AI, and generative outputs — text, images, audio and video — must carry visible labels plus machine‑readable provenance markers.
• Models already on the market before August 2 have until December 2, 2026 to add machine‑readable markings, and the EU AI Office can now request technical documents, run inspections and demand remedial steps.
• The Office can restrict or remove offerings and levy large fines — typically up to 3% of global turnover or €15 million, with more severe breaches subject to higher penalties — to enforce compliance.
• The law uses a tiered risk approach that bans clearly unacceptable practices like social scoring and real‑time biometric ID in public, requires strict safeguards for high‑risk systems, and applies to non‑EU firms that place models on the European market.