This study, presented by Chang Shi, et al., introduces Fentanyl-Hunter, a new system that finds and identifies fentanyl compounds and their changed forms using machine learning and molecular networks. To do this, researchers built a machine learning tool (Fentanyl_Finder) by training it with many fentanyl and non-fentanyl mass spectrometry readings, and they also created a multi-layered molecular network (Fentanyl_ID) that uses how similar molecules look in MS and how far apart their masses are to figure out their structures. Fentanyl-Hunter's classification model was very accurate, scoring 0.868 ± 0.02, and its molecular network could identify over 87% of known fentanyls. The system found 35 changed forms (metabolites) from four common fentanyl types during how the body processes them. It also looked back at over 605,000 MS files from public databases and found fentanyl, sufentanil, norfentanyl, or remifentanil acid in more than 250 samples from eight major countries, showing that fentanyl might be much more common than expected. The study concluded that Fentanyl-Hunter makes analysis more efficient and finds more fentanyl-related substances than older methods. It also noted that using AI in this way for drug regulation is safe because it doesn't reveal the exact chemical structures or MS data of fentanyls directly.