EU AI Labeling rules: Transparency Is not the same as Trust

Yesterday, the EU's new AI transparency obligations became applicable under Article 50 of the AI Act, requiring certain AI-generated or AI-manipulated content to be labelled and users to be informed when they interact with AI systems. The objective is understandable and, in many cases, justified: people should not be deceived. Deepfakes designed to mislead, impersonate or manipulate deserve transparency measures. Trust matters.

But Europe risks repeating a familiar mistake. Over the past decade, consumers have been inundated with privacy notices, cookie banners and consent requests. The result was not more informed citizens. It was fatigue. Many clicked through disclosures without reading them because there were simply too many. Today, we risk creating the same phenomenon with AI.

The new rules focus heavily on whether AI was used, rather than whether the content is actually misleading. A harmless image touched up by AI (you use AI to resize an image), a marketing text refined by a generative AI tool (you use a grammar check), or a video enhanced by AI (you use AI to improve cuts between segments) may require disclosures, while misleading or deceptive content created entirely by humans can continue circulating without any equivalent label. 

That raises an important question: what problem are we trying to solve?

Consumers care about whether content is truthful, fair and reliable. Most do not care whether a sentence was drafted by a human, assisted by AI, checked by AI, or translated by AI. Yet the law increasingly treats these distinctions as decisive. The result may be a growing number of labels that provide little meaningful information while drawing regulatory attention away from genuinely deceptive practices.

The challenge becomes even greater because AI use exists on a spectrum. Content may be entirely generated by AI, lightly edited by AI, or simply checked by an AI grammar tool. The legislation struggles to reflect this granularity. A (nearly) binary "AI-generated" label cannot capture the reality of how modern digital content is created. It risks implying a level of risk or artificiality that does not correspond to the actual use of technology.

Europe should also consider the broader economic impact. Compliance costs are not borne equally. Large technology companies can absorb new labelling, monitoring and governance requirements. Smaller businesses, publishers, creative professionals and SMEs often cannot. Every additional disclosure obligation may appear modest in isolation, but collectively they add to the growing regulatory burden already facing European innovators. Furthermore, disproportionate requirements may discourage beneficial uses of AI, even where the risks are low, which in turn is not beneficial for the EU economy as a whole.

At a time when European policymakers are rightly discussing competitiveness, simplification and innovation, we should be careful not to create rules that discourage the responsible use of privacy-, productivity-, and innovation-enhancing technologies. Europe does not become more trustworthy by regulating benign AI-generated content differently than comparable human-created content. It becomes more bureaucratic.

The ultimate question should not be: Was AI involved? It should be: Is the content misleading?

Until European regulation focuses more on outcomes and harms than on the mere presence of AI, we risk overwhelming consumers with labels, diluting trust, and making innovation harder without making society safer. The challenge is not detecting AI everywhere. The challenge is tackling deception wherever it occurs.

Europe needs a smarter approach. Rather than obsessing over whether AI was involved in the creation of content, policymakers should focus on whether consumers are actually being deceived or harmed. Transparency should be risk-based, meaningful and proportionate, not a blanket obligation attached to every instance of AI use. 

As the European Commission pursues its competitiveness and simplification agenda, it should revisit these requirements with fresh eyes. The goal should be to tackle deception, fraud and manipulation regardless of the technology used, while allowing businesses and citizens to benefit from AI responsibly. Europe cannot afford to regulate itself into irrelevance. If we want trust, innovation and growth, we must stop treating AI as the problem and focus on preventing harm and promoting integrity. 

Because transparency should be a means to an end, not an end in itself.