Overview
- The bulk of the EU AI Act's obligations entered into force on Sunday, requiring providers to tell users when they are interacting with AI and to attach machine‑readable metadata to generated or altered content.
- Mandatory visible watermarking for synthetic images, audio and video has been postponed until 2 December 2026 while tougher controls for high‑risk systems remain scheduled for December 2027 and August 2028.
- The Commission has expanded enforcement capacity by staffing the EU AI Office with specialists and launching confidential whistleblower and user reporting channels to collect evidence and open probes under the new law.
- Companies now face urgent compliance work such as taking inventories of AI use, assigning responsibilities, reviewing suppliers and strengthening cyber controls because failures carry heavy fines and market or reputational consequences.
- The AI Act combines specific bans on unacceptable‑risk systems with phased deadlines to let businesses adapt, a design that industry observers say will push non‑EU firms to follow EU rules and shape global AI practice.