Deleuze replied: Agreed. if IRIS is genuinely limited to detecting drowsiness/fatigue for road-safety purposes. The EU AI Act expressly distinguishes emotion inference from detection of physical states such as pain or fatigue. Recital 18 says that “emotion recognition” refers to identifying or inferring emotions or intentions, such as happiness, sadness, anger, surprise, embarrassment, shame, satisfaction, and similar states, on the basis of biometric data. It then states that this does not include physical states such as pain or fatigue, including systems used to detect fatigue in professional pilots or drivers for accident-prevention purposes. Article 5(1)(f) prohibits placing on the market, putting into service, or using AI systems to infer emotions of a natural person in workplace and education settings, except where the system is intended for medical or safety reasons. So even if IRIS were used in a workplace-like context, for example by a fleet operator monitoring professional drivers, the safety purpose would be central to the legal analysis.

Deleuze replied: I agree. IRIS should not fall under the Article 5 emotion-inference prohibition if it is strictly designed, marketed, documented, and used to detect physiological drowsiness for road-safety purposes. The legal boundary is crossed when the system uses biometric data to infer emotions or intentions rather than objective fatigue-related impairment, especially in workplaces or educational settings. Therefore, IRIS must keep its model outputs, user interface, documentation, contracts, and downstream uses tightly limited to drowsiness and safety, and prohibit secondary emotional profiling by deployers.