Bridging AI advancements with risk assessment needs: A journey towards effective use and regulatory acceptance

In recent years, the development of artificial intelligence (AI), including machine learning (ML) and deep learning (DL), has not only captured the attention of scientists, industry, and society due to its vast potential, but also sparked debates regarding the reliability, transparency, and legal as well as ethical implications of AI systems and tools. At the same time, toxicology is undergoing a paradigm shift toward New Approach Methodologies (NAMs). The implementation of NAMs necessitates an evolution of chemical risk assessment (CRA) approaches, giving rise to the concept of next-generation risk assessment (NGRA). It also transforms the field of regulatory toxicology into a data-rich discipline entailing the generation and management of increasingly large and complex datasets. For data handling and interpretation, these changes require effective integration of AI tools in chemical testing and assessment. In this narrative review, we examine the current landscape of AI applications relevant to CRA and NGRA. These tools can be broadly categorized into two domains: evidence management and evidence generation. They might support the discovery of toxicological mechanisms, biomarker identification, Adverse Outcome Pathway (AOP) development, and modeling of toxicokinetic and toxicodynamic processes as well as exposure assessment. Finally, we identify areas where further progress is required to ensure the successful integration of AI tools into CRA by extending beyond technical validation and guiding the development of detailed criteria for their regulatory acceptance.