On August 2nd, the European Union began requiring companies to label AI systems they deploy in public—disclosing whether a chatbot is machine-generated, whether a video is synthetic, whether a decision about you was made by an algorithm.
This is supposed to make the invisible legible and give people information they can actually use to change their behavior. It will not.
We have known this since 1992, when the EU mandated energy efficiency ratings for all major appliances. Refrigerators, washing machines, air conditioners would now carry a letter grade from A to G, displayed in a standardized label.
The logic was frictionless. Show people how much energy a device consumes. People will choose the efficient option, manufacturers will compete on efficiency, and energy use will drop. What happened instead is the mechanism's hidden fault. Once a label exists, it becomes a category to game. Manufacturers learned to design products that performed well on the test that determined the rating, then differently in the home. They reorganized product lines so even mediocre models could claim A ratings by measuring against narrower product classes.
Disclosure and behavioral change are not the same mechanism.
The EU AI transparency labels face the identical structural problem. Companies will disclose what regulators require because non-compliance carries fines—but disclosure and behavioral change are not the same mechanism. A user who learns an AI system generated a video still faces the original choice that made them seek the video. They must decide whether to view it as evidence of something they wanted to believe or as entertainment at face value. The label satisfied the regulatory obligation. It allows both the company and the user to feel that due diligence has occurred.
Watch what happens in your own practice when someone hands you a tool with a "transparent" purpose. You'll use it for the purpose you already had. The label made you feel informed. It did not change what you chose.