The EU's labeling requirement, while well-intentioned, raises serious practical concerns. For small and medium-sized businesses, the cost of implementing robust labeling systems can be prohibitive. A local news aggregator using AI to summarise stories must now invest in metadata infrastructure, diverting resources from core journalism.
Another problem is definitional. What counts as 'AI-generated'? Many tools assist human creators without fully automating content. An AI-suggested phrase in a reporter's article may not warrant a label, but drawing that line is tricky. Over-labeling could desensitise users, while under-labeling invites fines.
Furthermore, the rule faces enforcement challenges. Much AI-generated content comes from outside the EU, where companies may ignore the mandate. The Commission will rely on platform providers to police their feeds, but those same platforms profit from viral synthetic content. Without rigorous cross-border cooperation, the law will have limited effect.
Critics also warn of unintended chilling effects. Artists and hobbyists using AI tools for fun or satire may be caught by broad definitions. The exemption for satire only helps if clear guidance exists, which it does not yet. Until that clarity arrives, some creators may avoid testing novel uses of AI altogether.
The mandate may become a bureaucratic burden that hinders innovation without achieving its core goal. A more targeted approach, focusing on high-risk deepfakes and political ads, might have been more effective and less disruptive.