História
agosto 4, 2026
EU AI Labels Arrive, but the Real Compliance Deadline Is Still Moving
Europe’s new AI transparency regime has begun, forcing chatbots and synthetic media toward clearer disclosures. Yet legacy-system grace periods, postponed high-risk rules and split enforcement mean the rollout will be far from uniform.
Europe’s push to make AI visible has moved from policy ambition to enforcement. But for companies operating across the bloc, the rules have landed with a thicket of different clocks attached.
On August 2, the EU’s AI Act transparency obligations began applying, requiring firms to tell users when they are dealing with AI and to mark synthetic or manipulated material in machine-readable form. Platforms using realistic AI-generated images, audio or video must label deepfakes, while providers must make chatbot disclosures unless the artificial nature of the interaction is obvious. The Commission’s rationale is blunt: people should be able to judge what they are seeing, and how much to trust it. “People should know when they are interacting with AI or exposed to AI-generated content.”1
The rollout has been cast as AI’s “cookie banner” moment—a new layer of disclosure that could become routine across the internet.2 Companies that fail to comply face potential penalties of up to €15 million or 3% of global annual turnover. Yet the apparent August deadline is not universal: systems launched before that date were initially described as having a four-month window to catch up.1
The compliance picture became still more complicated as the Digital Omnibus delayed parts of the Act. Watermarking duties for systems already on the market now have until December 2, 2026; enforcement for two categories of high-risk AI systems has been pushed to 2027 and 2028. Enforcement is also divided among the EU AI Office, national authorities and the European Data Protection Supervisor, depending on the system and user.3
Meta, meanwhile, has signed the voluntary transparency code ahead of the obligations, while promising to prevent a “growing array of different labels and disclosures” from confusing users and regulators. The irony is hard to miss: the company made that pledge after rolling out its own bespoke labeling system.4
The EU has supplied optional common icons, hoping to curb that fragmentation. The labels may be optional; the obligation to disclose is not.1