In Table 11, the row for “Skin sensitization prediction” references Zhang et al. [20]. However, reference [20] is a study on odor prediction and aroma mixture design, not skin sensitization. Is this a citation error? If not, please justify why a paper on fragrance design is used to support performance metrics for a skin sensitization prediction model.
The paper suggests that AI models can support compliance with EU regulations and replace animal testing, yet no AI-driven toxicology model is currently approved by the EU or U.S. FDA for standalone safety assessment of cosmetics. On what basis do the authors claim that these in silico models are ready for regulatory adoption, given the absence of formal regulatory validation frameworks for AI in cosmetic safety?
In the Discussion, the authors state that “AI can accelerate the identification of safer, more sustainable ingredients, such as biodegradable polymers,” but no supporting reference is provided in that paragraph. Which specific studies or results demonstrate AI’s successful acceleration of biodegradable polymer discovery for cosmetics? Please cite the relevant evidence.