Skincare products, which provide cleansing, protection, moisturization, and appearance enhancement, represent a substantial segment of the global cosmetic industry. However, their widespread use has been associated with various adverse effects, including allergic reactions, irritation, phototoxicity, and even systemic toxicity, raising significant concerns regarding consumer safety. Traditional safety assessment relies largely on in vivo and in vitro models, which, although informative, are constrained by high costs, lengthy testing timelines, and ethical considerations. In response to these limitations, computational and in silico approaches have emerged as powerful alternatives for toxicological evaluation. These methods encompass artificial intelligence (AI) and machine learning (ML) techniques, quantitative structure–activity relationship (QSAR) models, mechanistic frameworks such as adverse outcome pathways (AOPs), rule-based expert systems, and read-across strategies. This review critically examines the current landscape of predictive modelling for skincare safety, highlighting both the opportunities and the challenges associated with integrating computational approaches into cosmetic safety assessment. By enabling greater predictive accuracy, reducing reliance on animal testing, and accelerating product development, in silico methodologies offer a transformative pathway toward safer, more effective skincare products. AI- and ML-based models demonstrate strong potential for capturing complex, nonlinear relationships across large and heterogeneous datasets. Nevertheless, key challenges, such as model interpretability, overfitting risk, and regulatory acceptance, must be systematically addressed to establish these predictive tools as reliable and practical components of next-generation skincare safety evaluation. • Overview of AI/ML and mechanistic methods for skincare safety prediction. • Traditional testing is limited by cost, time, and ethics. • Prediction is enhanced by the in-silico tools (AI, ML, QSAR, AOPs). • Assessment of the benefits, limitations, and regulatory implications of in-silico tools. • Challenges remain on model transparency, robustness, and regulatory approvals.
Qayyum et al. (Wed,) studied this question.