Abstract
The EU Artificial Intelligence Act (AI Act) is the world's first comprehensive, cross-sectoral legal framework dedicated specifically to AI. It introduces a structured regulatory approach to ensure that AI systems are safe, transparent, and trustworthy. To foster innovation, it includes research exemptions that place certain AI systems - those under development or used solely for scientific research - outside of its scope and obligations. However, this paper argues that these exemptions rely on distinctions that may not fully capture the realities of contemporary AI research. These include the unclear divide between research and commercial activities, and between lab-based development and real-world testing. Through legal analysis and practical scenarios, we demonstrate how the blurred boundaries between academic and commercial interests, as well as between controlled research and real-world use, create regulatory uncertainty and open the door to potential misuse. The paper highlights the risks stemming from vague definitions and the lack of harmonized guidance. It ultimately calls for clearer guidance, stronger safeguards, and more realistic frameworks that reflect the complexities of modern AI research.