Forensic toxicology has long been tasked with addressing fundamental questions of causation in medico-legal investigations, such as whether death resulted from poisoning or drug use. Although advanced analytical platforms, including liquid chromatography–tandem mass spectrometry (LC–MS/MS) and gas chromatography-mass spectrometry (GC–MS), provide highly sensitive and specific data, the interpretation of these complex datasets remains a significant challenge. In recent years, artificial intelligence (AI) has emerged as a transformative tool in this domain, offering not only enhanced analytical speed but also the capacity to generate deeper, data-driven insights into toxicological findings. This review critically examines the application of AI and machine learning techniques within forensic toxicology.
Key areas of focus include predictive toxicology, the development of AI-driven spectral libraries, automation of analytical workflows, and the integration of multi-omics data for comprehensive toxicological profiling. Furthermore, the review discusses the current challenges of ensuring robustness, transparency, and admissibility of AI-derived evidence in legal contexts, and outlines potential future directions for the incorporation of AI in forensic practice.
Introduction
Forensic toxicology, a critical field at the nexus of science and justice, faces increasing complexity due to the rise of novel psychoactive substances, environmental toxins, and diverse poisons. Traditional analytical methods like GC-MS and LC-MS/MS generate vast, complex data sets, challenging human interpretation. Artificial Intelligence (AI) and machine learning (ML) have emerged as transformative tools, enabling rapid analysis, pattern recognition, and prediction of toxic effects.
AI applications in forensic toxicology include:
Data Analysis: ML models classify toxins and unknown compounds with improved accuracy.
Plant Toxins: High-resolution mass spectrometry combined with AI helps detect rare plant poisons in complex biological samples.
Postmortem Toxicology: AI models assist in interpreting toxin redistribution and degradation after death, improving case accuracy.
Pesticide Toxicity: AI accelerates identification of mixed pesticide exposures in biological samples.
Novel Psychoactive Substances (NPS): AI adapts to rapidly evolving drug landscapes, supporting proactive detection and monitoring.
Emerging trends involve integrating chemometrics with AI, automating spectral library curation, predictive toxicology modeling ("future poisoning"), digital twin biological systems, AI-assisted histopathology, and multi-omics data integration for comprehensive toxic risk assessment.
Practical forensic applications include spectrum deconvolution, rapid drug/poison classification, natural language processing for case data, and automation through AI-guided robotics.
Challenges remain, notably data quality and standardization across labs, the need for transparent and explainable AI ("black box" issue), costs, and legal acceptance. Efforts focus on collaborative data sharing and developing interpretable AI models to ensure forensic validity and courtroom trust.
Conclusion
AI is not intended to substitute forensic toxicologists; it is intended to empower forensic toxicologists. By augmenting the data surcharge burden, AI actually gives forensic toxicologists much more time to spend on what forensic toxicologists think is important, which is interpreting the results in the context of justice and human life. The future forensic toxic science will likely be hybrid, where human expertise and an AI in mind are working in tandem. They might work together to provide more rapid, more accurate and more accessible toxicology testing for both science and society.
References
[1] Di Giorgi, S. Pichini, F. P. Busardò, and G. Basile, “Artificial intelligence in new psychoactive substances analysis: state-of-the-art and future perspectives,” J Anal Toxicol, Jul. 2025, doi: 10.1093/JAT/BKAF071.
[2] K. Jurowski and A. Kro?niak, “Prediction of key toxicity endpoints of AP-238, a new psychoactive substance for clinical toxicology and forensic purposes using in silico methods,” Sci Rep, vol. 14, no. 1, pp. 1–14, Dec. 2024, doi: 10.1038/S41598-024-79453-5; SUBJMETA.
[3] Z. J. Zhu et al., “Liquid Chromatography Quadrupole Time-of-Flight Characterization of Metabolites Guided by the METLIN Database,” Nat Protoc, vol. 8, no. 3, p. 451, Feb. 2013, doi: 10.1038/NPROT.2013.004.
[4] H. Yu, J. Ding, T. Shen, M. Liu, Y. Li, and O. Fiehn, “MassCube improves accuracy for metabolomics data processing from raw files to phenotype classifiers,” Nature Communications , vol. 16, no. 1, pp. 1–15, Dec. 2025, doi: 10.1038/S41467-025-60640-5;TECHMETA.
[5] F. Wu, X. Zhang, Z. Fang, and X. Yu, “Support Vector Machine-Based Global Classification Model of the Toxicity of Organic Compounds to Vibrio fischeri,” Molecules, vol. 28, no. 6, p. 2703, Mar. 2023, doi: 10.3390/MOLECULES28062703/S1.
[6] E. Sisco and T. P. Forbes, “Forensic applications of DART-MS: A review of recent literature,” Forensic Chemistry, vol. 22, p. 100294, Mar. 2021, doi: 10.1016/J.FORC.2020.100294.
[7] A. Zissler, W. Stoiber, J. Geissenberger, P. Steinbacher, F. C. Monticelli, and S. Pittner, “Influencing factors on postmortem protein degradation for pmi estimation: A systematic review,” Diagnostics, vol. 11, no. 7, p. 1146, Jul. 2021, doi: 10.3390/DIAGNOSTICS11071146/S1.
[8] A. Di Giorgi, S. Pichini, F. P. Busardò, and G. Basile, “Artificial intelligence (AI) in New Psychoactive Substances (NPS) analysis: state-of-art and future perspectives,” J Anal Toxicol, Jul. 2025, doi: 10.1093/JAT/BKAF071.
[9] A. Gangwal and A. Lavecchia, “Artificial intelligence in preclinical research: enhancing digital twins and organ-on-chip to reduce animal testing,” Drug Discov Today, vol. 30, no. 5, p. 104360, May 2025, doi: 10.1016/J.DRUDIS.2025.104360.
[10] Q. He et al., “Dear-DIAXMBD: Deep Autoencoder Enables Deconvolution of Data-Independent Acquisition Proteomics,” Research, vol. 6, p. 0179, Jan. 2023, doi: 10.34133/RESEARCH.0179.
[11] V. Notarstefano et al., “A new approach to assess post-mortem interval: A machine learning-assisted label-free ATR-FTIR analysis of human vitreous humor,” Spectrochim Acta A Mol Biomol Spectrosc, vol. 327, p. 125326, Feb. 2025, doi: 10.1016/J.SAA.2024.125326
[12] F. Alam, T. S. Mohammed Alnazzawi, R. Mehmood, and A. Al-maghthawi, “A Review of the Applications, Benefits, and Challenges of Generative AI for Sustainable Toxicology,” Curr Res Toxicol, vol. 8, p. 100232, Jan. 2025, doi: 10.1016/J.CRTOX.2025.100232.