Europeans Are About to Find Out How Entrenched AI Is in Their Daily Lives

When you use online customer service within the EU, there may not be a real person on the other end; when you come across a political speech video, it could be a deepfake; even the article you are reading right now may have been generated with AI assistance. From now on, all of these situations could trigger mandatory "AI disclosure." Yet, as the transparency provisions of the EU's Artificial Intelligence Act take full effect, a brand-new concern is emerging: will people develop "disclosure fatigue" toward AI notices that are everywhere, and eventually just tune them out?

Unprecedented Transparency Obligations

The EU's Artificial Intelligence Act is the world's first comprehensive law regulating AI. Its transparency requirements clearly stipulate that when users interact with AI systems or view content generated or edited by AI, the relevant parties must inform users in a clear and unambiguous manner. This means chatbots must identify themselves, deepfake content must be labeled, and images, videos, and audio altered by AI can no longer remain "invisible." The regulators' original intent is to prevent AI from misleading the public and to ensure that people have the right to know about machines' actions.

This obligation has broad coverage, encompassing not only general-purpose systems like ChatGPT but also vertical scenarios such as medical diagnosis, recruitment screening, and customer service. Under the rules set to take effect, any organization deploying AI—whether a tech giant or a small startup—must establish robust disclosure mechanisms. Compliance costs, interface design, and user education will all become new challenges.

The Hidden Concern of "Disclosure Fatigue"

However, even supporters of the rules must remain alert to reality. Cookie banners on social media have already trained ordinary users to habitually click "Agree," and the emerging AI disclosure notices face the same risk. A study from the University of Cambridge notes that when people frequently encounter warnings and prompts, their brains gradually become less sensitive to such information. This "prompt fatigue" may cause truly important warnings to be drowned out in the noise. For AI disclosure, if every chat session pops up with "You are conversing with an AI," users may quickly come to see it as a routine distraction and forget to pay attention to genuinely dangerous deepfakes or manipulative content.

One cannot simply assume that "adding a label" equals protecting users. The form, timing, and frequency of disclosure are equally decisive to its effectiveness.

Moreover, cross-platform and cross-device scenarios make consistency a difficult problem. When an AI-generated video is shared across social platforms, the original publisher may have labeled it, but the reposted version can lose that information. Practical loopholes like these cast doubt on how effectively the new rules can be implemented.

The Tech Industry's Dilemma

For large technology companies, this may amount to little more than an interface update: adding a line of notice to a dialog box or embedding an icon in the corner of a media player. But the genuinely thorny part is the generative AI ecosystem—every link in the chain, from user-generated content and third-party plugins to API call sequences, may be required to pass along disclosure information. Many developers worry that product experiences will become fragmented in the name of compliance. Smaller AI companies are even more anxious about cost pressures, since building traceability systems and metadata solutions that meet EU standards requires substantial engineering and legal resources.

In the meantime, the industry is also exploring smarter ways forward. Some vendors have proposed standardized unified markers—such as watermarks, icons, or hidden metadata—so that disclosure is embedded in the content itself rather than delivered through explicit pop-ups each time. Others are developing context-aware disclosure methods that interrupt users only when necessary, trying to find a balance between transparency and experience.

Editor's Note: Transparency Is Not the End; Understanding Is

The EU's attempt deserves recognition: it is the first time that transparency in AI interactions has been enshrined as a civic right in law. But transparency itself is only a means; its ultimate goal should be to help the public rationally understand the boundaries and limitations of AI. If "disclosure" merely becomes legal boilerplate or an obscure corner mark, then even with labels in place, it will achieve little.

Perhaps in the future, AI disclosure will appear on interactive interfaces the way nutrition labels appear on food packaging, requiring the joint effort of content producers, platform regulators, and users themselves. Only when "informing" is truly transformed into "understanding" can we avoid habitually clicking "Agree" once more amid the tide of information overload.

This article was adapted from WIRED.