Artificial intelligence is playing an increasing role in drug discovery, chemical research and toxicology. However, many machine learning models still work as “black boxes”, making their predictions difficult to interpret and sometimes difficult to use in industrial or regulatory contexts.