There are increasing efforts to apply new approaches methodlogies (NAMs), such as high-throughput screening (HTS), to chemical risk assessment. Traditional methods are not suitable for large-scale chemical screening, emphasizing the need for computational approaches such as artificial intelligence (AI). However, to apply AI-based toxicity predictions effectively in regulatory contexts, in vivo experimental validation is essential. The Adverse Outcome Pathway (AOP) framework provides mechanistic links between in vitro HTS data and in vivo toxicity endpoints, enabling explainable predictions. In this study, we developed an AOP-based, knowledge-based AI model to screen the developmental neurotoxicity (DNT) of consumer product chemicals and validated the predictions using high-throughput behavioral screening of Caenorhabditis elegans.