{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,28]],"date-time":"2026-09-28T02:55:11Z","timestamp":1790564111550,"version":"4.1.0"},"reference-count":148,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,9,28]],"date-time":"2026-09-28T00:00:00Z","timestamp":1790553600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Toxicol."],"abstract":"<jats:p>Artificial intelligence (AI), particularly in the form of large language models (LLMs), offers promising support for toxicological risk assessment (TRA), which integrates multiple lines of evidence to support health-based conclusions, including those used in regulatory decision-making. Effective use of LLMs in TRA requires a clear understanding of their strengths and limitations, especially in the context of regulatory expectations for transparency, reproducibility, and accountability. This work explores the emerging role of LLMs in TRA through three integrated components: an overview of LLMs in the context of TRA, an exemplar case study examining LLM-assisted derivation (mainly using ChatGPT) of the Permitted Daily Exposure (PDE) for acetyl tributyl citrate under the draft ICH Q3E guideline, and a workflow hypothesis for the integration of LLMs into TRA. Observations from the case study suggest that LLMs may support pattern recognition, targeted information retrieval, and rapid summarization of large bodies of text, facilitating literature triage, extraction of study details, and evidence summaries. At the same time, the case study highlights important limitations for LLMs. Observed variability in LLM outputs reflects inherent model stochastic behavior, model limitations, and the intrinsic complexity of TRA, where different experts may reasonably reach distinct, yet scientifically defensible conclusions based on the available evidence. The observations from the case study, interpreted in the context of the available literature, are used to inform a hypothesis that LLMs may be most effectively integrated into TRA through structured, modular, human-supervised workflows. This hypothesis reflects the view that the suitability of LLM-assisted TRA depends not only on LLM capabilities, but also on how LLMs are integrated into the assessment process to address both LLM variability and the interpretive nature of TRA. The proposed pragmatic workflow separates data search, information extraction, and integrated summarization while maintaining human-in-the-loop control over interpretation, expert judgment, and final conclusions.<\/jats:p>","DOI":"10.3389\/ftox.2026.1877128","type":"journal-article","created":{"date-parts":[[2026,9,28]],"date-time":"2026-09-28T02:07:44Z","timestamp":1790561264000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["A pragmatic workflow for integrating large language models into toxicological risk assessment"],"prefix":"10.3389","volume":"8","author":[{"given":"Arianna","family":"Bassan","sequence":"first","affiliation":[{"name":"Innovatune srl","place":["Padova, Italy"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Alexander","family":"Amberg","sequence":"additional","affiliation":[{"name":"Preclinical Safety, Sanofi","place":["Frankfurt, Germany"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Maria Letizia","family":"Barreca","sequence":"additional","affiliation":[{"name":"In Silico Drug Discovery Lab (I2D Lab), Department of Pharmaceutical Sciences, University of Perugia","place":["Perugia, Italy"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Paul M.","family":"Bradley","sequence":"additional","affiliation":[{"name":"Paul Bradley Consulting","place":["Royston, United Kingdom"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Frank","family":"Bringezu","sequence":"additional","affiliation":[{"name":"Merck Healthcare KGaA","place":["Darmstadt, Germany"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Joyce","family":"De Paula Souza","sequence":"additional","affiliation":[{"name":"Swiss Centre for Applied Human Toxicology, Department of Pharmaceutical Sciences, University of Basel","place":["Basel, Switzerland"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Corinne","family":"Guicheney","sequence":"additional","affiliation":[{"name":"Global Toxicology Services, KaLy-Cell, a Company of GBA Pharma","place":["Plobsheim, France"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Elena","family":"Lo Piparo","sequence":"additional","affiliation":[{"name":"Global Regulatory Affairs, Sciences & Strategy, FMC","place":["Geneva, Switzerland"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Simona","family":"Kovarich","sequence":"additional","affiliation":[{"name":"Global Regulatory & Computational Toxicology Team, Hazard Communication and Chemicals Regulations, Merck Life Science S.r.l.","place":["Milan, Italy"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Andrey","family":"Massarsky","sequence":"additional","affiliation":[{"name":"Stantec Consulting Services Inc.","place":["Irvine, CA, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Georgia","family":"Maundrill","sequence":"additional","affiliation":[{"name":"Medicines Research Centre, GSK","place":["Stevenage, United Kingdom"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lance","family":"Molnar","sequence":"additional","affiliation":[{"name":"Global Pharmacology and Toxicology, Viatris","place":["Morgantown, WV, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Cristina","family":"Parenti","sequence":"additional","affiliation":[{"name":"Environment and Health Department, Istituto Superiore di Sanit\u00e0","place":["Rome, Italy"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Patricia","family":"Parris","sequence":"additional","affiliation":[{"name":"Pfizer Worldwide Research, Development and Medical","place":["Sandwich, United Kingdom"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Manuela","family":"Pavan","sequence":"additional","affiliation":[{"name":"Innovatune srl","place":["Padova, Italy"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"John F.","family":"Reichard","sequence":"additional","affiliation":[{"name":"Department of Environmental and Public Health Sciences, University of Cincinnati","place":["Cincinnati, OH, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ron","family":"Steigerwalt","sequence":"additional","affiliation":[{"name":"Ron Steigerwalt Regulatory Tox Consulting, LLC","place":["Palm Springs, CA, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Olga","family":"Tcheremenskaia","sequence":"additional","affiliation":[{"name":"Environment and Health Department, Istituto Superiore di Sanit\u00e0","place":["Rome, Italy"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Kenneth","family":"Unice","sequence":"additional","affiliation":[{"name":"Strategic Health Sciences, TRC Companies","place":["Pittsburgh, PA, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Christopher J.","family":"Waine","sequence":"additional","affiliation":[{"name":"Bibra Toxicology Advice and Consulting","place":["Wallington, Surrey, United Kingdom"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Geraldine","family":"Whelan","sequence":"additional","affiliation":[{"name":"Medicines Research Centre, GSK","place":["Stevenage, United Kingdom"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Angela","family":"White","sequence":"additional","affiliation":[{"name":"Medicines Research Centre, GSK","place":["Stevenage, United Kingdom"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Glenn J.","family":"Myatt","sequence":"additional","affiliation":[{"name":"Informence Labs LLC","place":["St. Petersburg, FL, United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,9,28]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"28191","DOI":"10.1007\/s00521-025-11666-9","article-title":"A survey on retrieval-augmentation generation (RAG) models for healthcare applications","volume":"37","author":"Abo El-Enen","year":"2025","journal-title":"Neural Comput. and Applic"},{"key":"B2","doi-asserted-by":"publisher","first-page":"1568","DOI":"10.1057\/s41599-024-04044-8","article-title":"Trust in AI: progress, challenges, and future directions","volume":"11","author":"Afroogh","year":"2024","journal-title":"Humanit Soc. Sci. Commun."},{"key":"B3","doi-asserted-by":"publisher","first-page":"100232","DOI":"10.1016\/j.crtox.2025.100232","article-title":"A review of the applications, benefits, and challenges of generative AI for sustainable toxicology","volume":"8","author":"Alam","year":"2025","journal-title":"Curr. Res. Toxicol."},{"key":"B4","doi-asserted-by":"publisher","first-page":"110","DOI":"10.1186\/s12911-025-02944-6","article-title":"The role of explainable artificial intelligence in disease prediction: a systematic literature review and future research directions","volume":"25","author":"Alkhanbouli","year":"2025","journal-title":"BMC Med. Inf. Decis. Mak."},{"key":"B5","doi-asserted-by":"publisher","first-page":"310","DOI":"10.1186\/s12911-020-01332-6","article-title":"Explainability for artificial intelligence in healthcare: a multidisciplinary perspective","volume":"20","author":"Amann","year":"2020","journal-title":"BMC Med. Inf. Decis. Mak."},{"key":"B6","doi-asserted-by":"publisher","first-page":"1622292","DOI":"10.3389\/frai.2025.1622292","article-title":"Survey and analysis of hallucinations in large language models: attribution to prompting strategies or model behavior","volume":"8","author":"Anh-Hoang","year":"2025","journal-title":"Front. Artif. Intell."},{"key":"B7","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1177\/15353702211052280","article-title":"Emerging technologies and their impact on regulatory science","volume":"247","author":"Anklam","year":"2022","journal-title":"Exp. Biol. Med. (Maywood)"},{"key":"B8","doi-asserted-by":"publisher","first-page":"730","DOI":"10.1002\/etc.34","article-title":"Adverse outcome pathways: a conceptual framework to support ecotoxicology research and risk assessment","volume":"29","author":"Ankley","year":"2010","journal-title":"Environ. Toxicol. Chem."},{"key":"B9","article-title":"ANSES (Agence nationale de s\u00e9curit\u00e9 sanitaire de l\u2019alimentation, de l\u2019environnement et du travail)","volume-title":"Analysis of the Most Appropriate Risk Management Option (RMOA) Substance Name: Tributyl O-acetylcitrate (ATBC) EC Number: 201-067-0 CAS Number: 77-90-7","year":"2016"},{"key":"B10","article-title":"Models overview","year":"2026","journal-title":"Claude Docs"},{"key":"B11","unstructured":"AI model and API providers analysis | artificial analysis"},{"key":"B12","unstructured":"Artificial analysis long context reasoning benchmark leaderboard"},{"key":"B13","article-title":"GPQA diamond benchmark leaderboard","year":""},{"key":"B14","article-title":"IFBench benchmark leaderboard","year":""},{"key":"B15","unstructured":"Comparison of AI models across intelligence, performance, price | artificial analysis\n          \n          \n          2026"},{"key":"B16","doi-asserted-by":"publisher","first-page":"274","DOI":"10.1038\/s41746-025-01670-7","article-title":"A framework to assess clinical safety and hallucination rates of LLMs for medical text summarisation","volume":"8","author":"Asgari","year":"2025","journal-title":"npj Digit. Med."},{"key":"B17","doi-asserted-by":"publisher","first-page":"107197","DOI":"10.1016\/j.infsof.2023.107197","article-title":"Transparency and explainability of AI systems: from ethical guidelines to requirements","volume":"159","author":"Balasubramaniam","year":"2023","journal-title":"Inf. Softw. Technol."},{"key":"B18","doi-asserted-by":"publisher","first-page":"e68998","DOI":"10.2196\/68998","article-title":"Scientific evidence for clinical text summarization using large language models: scoping review","volume":"27","author":"Bednarczyk","year":"2025","journal-title":"J. Med. Internet Res."},{"key":"B19","article-title":"On the opportunities and risks of foundation models","author":"Bommasani","year":"2021"},{"key":"B20","article-title":"The foundation model transparency index","author":"Bommasani","year":"2023"},{"key":"B21","doi-asserted-by":"publisher","first-page":"525","DOI":"10.1038\/s42256-024-00832-8","article-title":"Augmenting large language models with chemistry tools","volume":"6","author":"Bran","year":"2024","journal-title":"Nat. Mach. Intell."},{"key":"B22","doi-asserted-by":"publisher","first-page":"e0319159","DOI":"10.1371\/journal.pone.0319159","article-title":"Prompt architecture induces methodological artifacts in large language models","volume":"20","author":"Brucks","year":"2025","journal-title":"PLOS ONE"},{"key":"B23","article-title":"Principled instructions are all you need for questioning LLaMA-1\/2, GPT-3.5\/4","author":"Bsharat","year":"2023"},{"key":"B24","doi-asserted-by":"publisher","first-page":"26","DOI":"10.1038\/s43856-024-00717-2","article-title":"Current applications and challenges in large language models for patient care: a systematic review","volume":"5","author":"Busch","year":"2025","journal-title":"Commun. Med."},{"key":"B25","doi-asserted-by":"publisher","first-page":"100632","DOI":"10.1016\/j.cosrev.2024.100632","article-title":"A contemporary review on chatbots, AI-powered virtual conversational agents, ChatGPT: applications, open challenges and future research directions","volume":"52","author":"Casheekar","year":"2024","journal-title":"Comput. Sci. Rev."},{"key":"B26","doi-asserted-by":"publisher","first-page":"20","DOI":"10.1038\/538020a","article-title":"Can we open the black box of AI?","volume":"538","author":"Castelvecchi","year":"2016","journal-title":"Nat. News"},{"key":"B27","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3641289","article-title":"A survey on evaluation of large language models","volume":"15","author":"Chang","year":"2024","journal-title":"ACM Trans. Intell. Syst. Technol."},{"key":"B28","unstructured":"Cosmetic ingredient review expert panel. Safety assessment of acetyl trialkyl citrates as used in cosmetics. Cosmetic ingredient review expert panel re-review\n          \n          2019"},{"key":"B29","volume-title":"Toxicity Review for Acetyl Tri-n-Butyl Citrate (ATBC)\u201d, Prepared for the U.S. Consumer Product Safety Commission (CPSC)","year":"2018"},{"key":"B30","doi-asserted-by":"crossref","first-page":"287","DOI":"10.1016\/B978-0-12-813697-3.00031-7","article-title":"Computational methods to predict toxicity","volume-title":"The History of Alternative Test Methods in Toxicology","author":"Cronin","year":"2019"},{"key":"B31","unstructured":"Artificial intelligence (AI) taxonomy\n          \n          \n          2025"},{"key":"B32","article-title":"A complete survey on LLM-based AI chatbots","author":"Dam","year":"2024"},{"key":"B33","doi-asserted-by":"publisher","first-page":"517","DOI":"10.3390\/info16070517","article-title":"Mitigating LLM hallucinations using a multi-agent framework","volume":"16","author":"Darwish","year":"2025","journal-title":"Information"},{"key":"B34","doi-asserted-by":"publisher","first-page":"20","DOI":"10.1016\/j.comtox.2018.06.001","article-title":"Principles underpinning the use of new methodologies in the risk assessment of cosmetic ingredients","volume":"7","author":"Dent","year":"2018","journal-title":"Comput. Toxicol."},{"key":"B35","doi-asserted-by":"publisher","first-page":"100395","DOI":"10.1016\/j.comtox.2025.100395","article-title":"O-QT assistant: a multi-agent AI system for streamlined chemical hazard assessment and read-across analysis using the OECD QSAR toolbox API","volume":"37","author":"Djidrovski","year":"2025","journal-title":"Comput. Toxicol."},{"key":"B36","doi-asserted-by":"publisher","first-page":"110","DOI":"10.1145\/3596490","article-title":"Shortcut learning of large language models in natural language understanding","volume":"67","author":"Du","year":"2024","journal-title":"Commun. ACM"},{"key":"B37","unstructured":"Integrating AI into chemical safety assessment - flash report\n          \n          \n          2025"},{"key":"B38","doi-asserted-by":"publisher","first-page":"273a","DOI":"10.2903\/j.efsa.2005.273a","article-title":"Opinion of the scientific panel on food additives, flavourings, processing aids and materials in contact with food (AFC) related to a 10th list of substances for food contact materials","volume":"3","year":"2005","journal-title":"EFSA J."},{"key":"B39","unstructured":"AI@EFSA | EFSA\n          \n          \n          2025"},{"key":"B40","unstructured":"Reflection paper on the use of artificial intelligence (AI) in the medicinal product lifecycle EMA\/CHMP\/CVMP\/83833\/2023 committee for medicinal products for human use (CHMP) committee for medicinal products for veterinary use (CVMP)\n          \n          \n          2024"},{"key":"B41","unstructured":"Guiding principles of good ai practice in drug development\n          \n          \n          2026"},{"key":"B42","unstructured":"Regulation (EU) 2024\/1689 of the european parliament and of the council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending regulations (EC) no 300\/2008, (EU) no 167\/2013, (EU) no 168\/2013, (EU) 2018\/858, (EU) 2018\/1139 and (EU) 2019\/2144 and directives 2014\/90\/EU, (EU) 2016\/797 and (EU) 2020\/1828 (artificial intelligence act) (text with EEA relevance)\n          \n          \n          2024"},{"key":"B43","unstructured":"Ethics guidelines for trustworthy AI | shaping Europe\u2019s digital future\n          \n          \n          2025"},{"key":"B44","volume-title":"Domain-Driven Design: Tackling Complexity in the Heart of Software","author":"Evans","year":"2004"},{"key":"B45","doi-asserted-by":"crossref","DOI":"10.1145\/3637528.3671470","article-title":"A survey on RAG meeting LLMs: towards retrieval-augmented large language models","author":"Fan","year":"2024"},{"key":"B46","doi-asserted-by":"publisher","first-page":"e39250","DOI":"10.1097\/MD.0000000000039250","article-title":"ChatGPT in medicine: a cross-disciplinary systematic review of chatgpt\u2019s (artificial intelligence) role in research, clinical practice, education, and patient interaction","volume":"103","author":"Fatima","year":"2024","journal-title":"Medicine"},{"key":"B47","unstructured":"Considerations for the use of artificial intelligence to support regulatory decision-making for drug and biological products\n          \n          \n          2025"},{"key":"B48","unstructured":"Guiding principles of good AI practice in drug development\n          \n          \n          2026"},{"key":"B49","first-page":"27699","article-title":"Mathematical capabilities of ChatGPT","volume-title":"Proceedings of the 37th International Conference on Neural Information Processing Systems","author":"Frieder","year":"2023"},{"key":"B50","article-title":"Retrieval augmented generation evaluation in the era of large language models: a comprehensive survey","author":"Gan","year":"2025"},{"key":"B51","doi-asserted-by":"crossref","DOI":"10.1145\/3787279.3787299","article-title":"Prompt sentiment: the catalyst for LLM change","author":"Gandhi","year":"2025"},{"key":"B52","doi-asserted-by":"publisher","first-page":"2149","DOI":"10.1007\/s00204-025-04286-8","article-title":"Building trust in the integration of artificial intelligence into chemical risk assessment: findings from the 2024 ECETOC workshop","volume":"100","author":"Gant","year":"2026","journal-title":"Arch. Toxicol."},{"key":"B53","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.2312.10997","article-title":"Retrieval-augmented generation for large language models: a survey","author":"Gao","year":"2023"},{"key":"B54","doi-asserted-by":"crossref","DOI":"10.24963\/ijcai.2025\/1073","article-title":"Resolving conflicting evidence in automated fact-checking: a study on retrieval-augmented LLMs","author":"Ge","year":"2025"},{"key":"B55","doi-asserted-by":"publisher","first-page":"665","DOI":"10.1038\/s42256-020-00257-z","article-title":"Shortcut learning in deep neural networks","volume":"2","author":"Geirhos","year":"2020","journal-title":"Nat. Mach. Intell."},{"key":"B56","article-title":"Gemini: a family of highly capable multimodal models","year":"2025"},{"key":"B57","doi-asserted-by":"publisher","first-page":"29","DOI":"10.1016\/0300-483X(95)03035-E","article-title":"Historical perspective on risk assessment in the federal government","volume":"102","author":"Graham","year":"1995","journal-title":"Toxicology"},{"key":"B58","article-title":"15th global summit on regulatory science (GSRS25)","volume-title":"Annual Conference Theme: Building a Strong Regulatory Science with Tomorrow\u2019s Technologies - September 15-17, 2025","year":"2025"},{"key":"B59","article-title":"Bias in large language models: origin, evaluation, and mitigation","author":"Guo","year":"2024"},{"key":"B60","doi-asserted-by":"publisher","first-page":"1269932","DOI":"10.3389\/frai.2023.1269932","article-title":"Artificial intelligence as the new frontier in chemical risk assessment","volume":"6","author":"Hartung","year":"","journal-title":"Front. Artif. Intell."},{"key":"B61","doi-asserted-by":"publisher","first-page":"559","DOI":"10.14573\/altex.2309191","article-title":"ToxAIcology - the evolving role of artificial intelligence in advancing toxicology and modernizing regulatory science","volume":"40","author":"Hartung","year":"","journal-title":"ALTEX - Altern. Animal Experimentation"},{"key":"B62","doi-asserted-by":"publisher","first-page":"1649155","DOI":"10.3389\/frai.2025.1649155","article-title":"AI, agentic models and lab automation for scientific discovery \u2014 the beginning of scAInce","volume":"8","author":"Hartung","year":"2025","journal-title":"Front. Artif. Intell."},{"key":"B63","doi-asserted-by":"publisher","first-page":"3065","DOI":"10.1007\/s00204-025-03978-5","article-title":"Assessing risk of bias in toxicological studies in the era of artificial intelligence","volume":"99","author":"Hartung","year":"","journal-title":"Arch. Toxicol."},{"key":"B64","doi-asserted-by":"publisher","first-page":"200","DOI":"10.1038\/s41746-025-01596-0","article-title":"Is regulatory science ready for artificial intelligence?","volume":"8","author":"Hartung","year":"","journal-title":"npj Digit. Med."},{"key":"B65","doi-asserted-by":"publisher","first-page":"104403","DOI":"10.1016\/j.yrtph.2019.104403","article-title":"Genetic toxicology in silico protocol","volume":"107","author":"Hasselgren","year":"2019","journal-title":"Regul. Toxicol. Pharmacol."},{"key":"B66","doi-asserted-by":"publisher","first-page":"4397","DOI":"10.1038\/s44319-025-00560-z","article-title":"The future of reviews: will LLMs render them obsolete?","volume":"26","author":"Hauser","year":"2025","journal-title":"EMBO Rep."},{"key":"B67","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2503.04739","article-title":"Responsible artificial intelligence systems: a roadmap to society\u2019s trust through trustworthy AI, auditability, accountability, and governance","author":"Herrera-Poyatos","year":"2025"},{"key":"B68","article-title":"The curious case of neural text degeneration","author":"Holtzman","year":"2019"},{"key":"B69","doi-asserted-by":"publisher","first-page":"e441","DOI":"10.1016\/S2589-7500(24)00111-0","article-title":"From text to treatment: the crucial role of validation for generative large language models in health care","volume":"6","author":"Hond","year":"2024","journal-title":"Lancet Digital Health"},{"key":"B70","article-title":"Understanding the planning of LLM agents: a survey","author":"Huang","year":"2024"},{"key":"B71","unstructured":"What is a chatbot? | IBM\n          \n          \n          2021"},{"key":"B72","volume-title":"Guideline for Extractables and Leachables Q3E, Draft Version, Endorsed on 01 August 2025","year":"2025"},{"key":"B73","unstructured":"ISO\/IEC 42001:2023\n          \n          \n          2023"},{"key":"B74","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3571730","article-title":"Survey of hallucination in natural language generation","volume":"55","author":"Ji","year":"2023","journal-title":"ACM Comput. Surv."},{"key":"B75","volume-title":"Mitigating LLM Hallucinations: A Comprehensive Review of Techniques and Architectures","author":"Joshi","year":"2025"},{"key":"B76","article-title":"Why language models hallucinate","author":"Kalai","year":"2025"},{"key":"B77","doi-asserted-by":"publisher","first-page":"187","DOI":"10.1038\/s41746-025-01519-z","article-title":"Retrieval augmented generation for 10 large language models and its generalizability in assessing medical fitness","volume":"8","author":"Ke","year":"2025","journal-title":"npj Digit. Med."},{"key":"B78","doi-asserted-by":"publisher","first-page":"735","DOI":"10.1007\/s00204-023-03666-2","article-title":"Artificial intelligence (AI)\u2014It\u2019s the end of the tox as we know it (and I feel fine)","volume":"98","author":"Kleinstreuer","year":"2024","journal-title":"Arch. Toxicol."},{"key":"B79","article-title":"Prompt-hacking: the new p-hacking?","author":"Kosch","year":"2025"},{"key":"B80","article-title":"LLLMs: a data-driven survey of evolving research on limitations of large language models","author":"Kostikova","year":"2025"},{"key":"B81","doi-asserted-by":"publisher","first-page":"667","DOI":"10.14573\/altex.2004041","article-title":"Development of an evidence-based risk assessment framework","volume":"39","author":"Krewski","year":"2022","journal-title":"ALTEX"},{"key":"B82","article-title":"Retrieval-augmented generation for knowledge-intensive NLP tasks","author":"Lewis","year":"2021"},{"key":"B83","doi-asserted-by":"publisher","first-page":"1379260","DOI":"10.3389\/fdsfr.2024.1379260","article-title":"Assessing the performance of large language models in literature screening for pharmacovigilance: a comparative study","volume":"4","author":"Li","year":"","journal-title":"Front. Drug Saf. Regul."},{"key":"B84","doi-asserted-by":"publisher","first-page":"108013","DOI":"10.1016\/j.cmpb.2024.108013","article-title":"ChatGPT in healthcare: a taxonomy and systematic review","volume":"245","author":"Li","year":"","journal-title":"Comput. Methods Programs Biomed."},{"key":"B85","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.2510.24476","article-title":"Mitigating hallucination in large language models (LLMs): an application-oriented survey on RAG, reasoning, and agentic systems","author":"Li","year":"2025"},{"key":"B86","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.2509.18970","article-title":"LLM-based agents suffer from hallucinations: a survey of taxonomy, methods, and directions","author":"Lin","year":"2025"},{"key":"B87","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.2305.18703","article-title":"Domain specialization as the key to make large language models disruptive: a comprehensive survey","author":"Ling","year":"2023"},{"key":"B88","article-title":"Lost in the middle: how language models use long contexts","author":"Liu","year":"2023"},{"key":"B89","doi-asserted-by":"publisher","first-page":"1","DOI":"10.3390\/biomedinformatics2010001","article-title":"Explainable artificial intelligence (XAI) in biomedicine: making AI decisions trustworthy for physicians and patients","volume":"2","author":"L\u00f6tsch","year":"2021","journal-title":"BioMedInformatics"},{"key":"B90","doi-asserted-by":"publisher","first-page":"51","DOI":"10.1007\/s40572-025-00514-6","article-title":"Navigating the AI frontier in toxicology: trends, trust, and transformation","volume":"12","author":"Luechtefeld","year":"2025","journal-title":"Curr. Envir Health Rpt"},{"key":"B91","doi-asserted-by":"publisher","first-page":"525","DOI":"10.1021\/acs.chemrestox.4c00050","article-title":"Deus Ex machina? The rise of artificial intelligence in toxicology","volume":"37","author":"Lui","year":"2024","journal-title":"Chem. Res. Toxicol."},{"key":"B92","doi-asserted-by":"publisher","first-page":"110","DOI":"10.1002\/em.20369","article-title":"Recent developments in frameworks to consider human relevance of hypothesized modes of action for tumours in animals","volume":"49","author":"Meek","year":"2008","journal-title":"Environ Mol Mutagen"},{"key":"B93","article-title":"Model cards for model reporting","author":"Mitchell","year":"2018"},{"key":"B94","doi-asserted-by":"publisher","first-page":"e70047","DOI":"10.1002\/wcms.70047","article-title":"Automating data extraction from scientific literature and general PDF files using large language models and KNIME: an application in toxicology","volume":"15","author":"Moreira-Filho","year":"2025","journal-title":"WIREs Comput. Mol. Sci."},{"key":"B95","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1016\/j.yrtph.2018.04.014","article-title":"In silico toxicology protocols","volume":"96","author":"Myatt","year":"2018","journal-title":"Regul. Toxicol. Pharmacol."},{"key":"B96","doi-asserted-by":"publisher","first-page":"671","DOI":"10.1038\/d41586-023-02366-2","article-title":"ChatGPT is a black box: how AI research can break it open","volume":"619","year":"2023","journal-title":"Nature"},{"key":"B97","volume-title":"Artificial Intelligence Risk Management Framework (AI RMF 1.0). Gaithersburg, MD","year":"2023"},{"key":"B98","volume-title":"Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile. Gaithersburg, MD","year":"2024"},{"key":"B99","unstructured":"AI RMF playbook\n          \n          \n          2026"},{"key":"B100","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2507.18910","article-title":"A systematic review of key retrieval-augmented generation (RAG) systems: progress, gaps, and future directions","author":"Oche","year":"2025"},{"key":"B101","unstructured":"Revised guidance document on developing and assessing adverse outcome pathways ENV\/JM\/MONO(2013)6. OECD environment, health and safety publications\n          \n          \n          2017"},{"key":"B102","volume-title":"Overview of Concepts and Available Guidance Related to Integrated Approaches to Testing and Assessment (IATA)","year":"2020"},{"key":"B103","article-title":"GPT-5 system card","year":"","journal-title":"OpenAI"},{"key":"B104","unstructured":"Introducing deep research"},{"key":"B105","unstructured":"Update to GPT-5 system card: GPT-5.2"},{"key":"B106","unstructured":"Models | OpenAI API\n          \n          \n          2026"},{"key":"B107","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.2203.02155","article-title":"Training language models to follow instructions with human feedback","author":"Ouyang","year":"2022"},{"key":"B108","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3697010","article-title":"An empirical study of the non-determinism of ChatGPT in code generation","volume":"34","author":"Ouyang","year":"2025","journal-title":"ACM Trans. Softw. Eng. Methodol."},{"key":"B109","doi-asserted-by":"publisher","first-page":"101885","DOI":"10.1016\/j.jsis.2024.101885","article-title":"Responsible artificial intelligence governance: a review and research framework","volume":"34","author":"Papagiannidis","year":"2025","journal-title":"J. Strategic Inf. Syst."},{"key":"B110","doi-asserted-by":"crossref","first-page":"6048","DOI":"10.18653\/v1\/2024.acl-long.329","article-title":"On measuring faithfulness or self-consistency of natural language explanations","volume-title":"Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","author":"Parcalabescu","year":"2024"},{"key":"B111","doi-asserted-by":"publisher","first-page":"30","DOI":"10.32384\/jeahil20623","article-title":"The role of ChatGPT in developing systematic literature searches: an evidence summary","volume":"20","author":"Parisi","year":"2024","journal-title":"J. Eur. Assoc. Health Info Libr."},{"key":"B112","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.2503.23037","article-title":"Agentic large language models, a survey","author":"Plaat","year":"2025"},{"key":"B113","doi-asserted-by":"publisher","first-page":"e0311937","DOI":"10.1371\/journal.pone.0311937","article-title":"Assessing the accuracy and consistency of answers by ChatGPT to questions regarding carbon monoxide poisoning","volume":"19","author":"Qiu","year":"2024","journal-title":"PLoS ONE"},{"key":"B114","volume-title":"Language Models are Unsupervised Multitask Learners","author":"Radford","year":"2019"},{"key":"B115","doi-asserted-by":"publisher","first-page":"668","DOI":"10.1002\/jat.3413","article-title":"Effects of oral exposure to the phthalate substitute acetyl tributyl citrate on female reproduction in mice","volume":"37","author":"Rasmussen","year":"2017","journal-title":"J. Appl. Toxicol."},{"key":"B116","doi-asserted-by":"publisher","first-page":"206","DOI":"10.1038\/s42256-019-0048-x","article-title":"Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead","volume":"1","author":"Rudin","year":"2019","journal-title":"Nat. Mach. Intell."},{"key":"B117","article-title":"Explainable artificial intelligence: understanding, visualizing and interpreting deep learning models","author":"Samek","year":"2017"},{"key":"B118","doi-asserted-by":"publisher","first-page":"158","DOI":"10.1007\/s42979-022-01043-x","article-title":"AI-Based modeling: techniques, applications and research issues towards automation, intelligent and smart systems","volume":"3","author":"Sarker","year":"2022","journal-title":"SN Comput. Sci."},{"key":"B119","doi-asserted-by":"publisher","first-page":"2546791","DOI":"10.1080\/2833373X.2025.2546791","article-title":"Large language models for data extraction in toxicology: implications and lessons learned from the clinical evidence domain","volume":"3","author":"Schmidt","year":"2025","journal-title":"Evidence-Based Toxicol."},{"key":"B120","doi-asserted-by":"publisher","first-page":"105649","DOI":"10.1016\/j.yrtph.2024.105649","article-title":"Comparison of permitted daily exposure (PDE) values for active pharmaceutical ingredients (APIs) - evidence of a robust approach","volume":"150","author":"Sehner","year":"2024","journal-title":"Regul. Toxicol. Pharmacol."},{"key":"B121","doi-asserted-by":"publisher","first-page":"31150","DOI":"10.1038\/s41598-024-82501-9","article-title":"Explainable AI improves task performance in human\u2013AI collaboration","volume":"14","author":"Senoner","year":"2024","journal-title":"Sci. Rep."},{"key":"B122","doi-asserted-by":"publisher","first-page":"404","DOI":"10.1186\/s12911-024-02824-5","article-title":"Assessing the accuracy and quality of artificial intelligence (AI) chatbot-generated responses in making patient-specific drug-therapy and healthcare-related decisions","volume":"24","author":"Shiferaw","year":"2024","journal-title":"BMC Med. Inf. Decis. Mak."},{"key":"B123","article-title":"OpenAI GPT-5 system card","author":"Singh","year":"2026"},{"key":"B124","doi-asserted-by":"publisher","first-page":"e0000388","DOI":"10.1371\/journal.pdig.0000388","article-title":"Hallucination or confabulation? Neuroanatomy as metaphor in large language models","volume":"2","author":"Smith","year":"2023","journal-title":"PLOS Digit. Health"},{"key":"B125","doi-asserted-by":"publisher","first-page":"153933","DOI":"10.1016\/j.tox.2024.153933","article-title":"Artificial intelligence-based data extraction for next generation risk assessment: is fine-tuning of a large language model worth the effort?","volume":"508","author":"Sonnenburg","year":"2024","journal-title":"Toxicology"},{"key":"B126","doi-asserted-by":"publisher","first-page":"1278","DOI":"10.1057\/s41599-024-03811-x","article-title":"AI hallucination: towards a comprehensive classification of distorted information in artificial intelligence-generated content","volume":"11","author":"Sun","year":"2024","journal-title":"Humanit Soc. Sci. Commun."},{"key":"B127","unstructured":"Framework conditions for the use of AI in medicinal product development and the regulatory process\n          \n          \n          2025"},{"key":"B128","doi-asserted-by":"publisher","first-page":"105388","DOI":"10.1016\/j.yrtph.2023.105388","article-title":"Artificial intelligence and real-world data for drug and food safety \u2013 a regulatory science perspective","volume":"140","author":"Thakkar","year":"2023","journal-title":"Regul. Toxicol. Pharmacol."},{"key":"B129","doi-asserted-by":"publisher","first-page":"104058","DOI":"10.1016\/j.drudis.2024.104058","article-title":"50 shades of AI in regulatory science","volume":"29","author":"Tong","year":"2024","journal-title":"Drug Discov. Today"},{"key":"B130","doi-asserted-by":"publisher","first-page":"1779","DOI":"10.1016\/S0140-6736(26)00603-3","article-title":"Fabricated citations: an audit across 2\u00b75 million biomedical papers","volume":"407","author":"Topaz","year":"2026","journal-title":"Lancet"},{"key":"B131","doi-asserted-by":"publisher","first-page":"2520764","DOI":"10.1080\/2833373x.2025.2520764","article-title":"Present and future of AI, open science, and transparency in regulatory science","volume":"3","author":"Tsaioun","year":"2025","journal-title":"Evidence-Based Toxicol."},{"key":"B132","doi-asserted-by":"publisher","first-page":"4","DOI":"10.21037\/jmai-21-25","article-title":"Factors influencing trust in medical artificial intelligence for healthcare professionals: a narrative review","volume":"5","author":"Tucci","year":"2022","journal-title":"J. Med. Artif. Intell."},{"key":"B133","volume-title":"Risk Assessment in the Federal Government: Managing the Process","year":"1983"},{"key":"B134","unstructured":"Toxicity forecasting (ToxCast)\n          \n          \n          2015"},{"key":"B135","volume-title":"Science and Decisions: Advancing Risk Assessment","year":"2009"},{"key":"B136","first-page":"6000","article-title":"Attention is all you need","volume-title":"Proceedings of the 31st International Conference on Neural Information Processing Systems","author":"Vaswani","year":"2017"},{"key":"B137","doi-asserted-by":"publisher","first-page":"312","DOI":"10.1093\/toxsci\/kfu199","article-title":"Adverse outcome pathway (AOP) development I: strategies and principles","volume":"142","author":"Villeneuve","year":"","journal-title":"Toxicol. Sci."},{"key":"B138","doi-asserted-by":"publisher","first-page":"321","DOI":"10.1093\/toxsci\/kfu200","article-title":"Adverse outcome pathway development II: best practices","volume":"142","author":"Villeneuve","year":"","journal-title":"Toxicol. Sci."},{"key":"B139","doi-asserted-by":"publisher","first-page":"41","DOI":"10.1038\/s41746-024-01029-4","article-title":"Prompt engineering in consistency and reliability with the evidence-based guideline for LLMs","volume":"7","author":"Wang","year":"2024","journal-title":"npj Digit. Med."},{"key":"B140","doi-asserted-by":"publisher","first-page":"509","DOI":"10.1038\/s41746-025-01840-7","article-title":"Accelerating clinical evidence synthesis with large language models","volume":"8","author":"Wang","year":"2025","journal-title":"npj Digit. Med."},{"key":"B141","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.2506.12594","article-title":"A comprehensive survey of deep research: systems, methodologies, and applications","author":"Xu","year":"2025"},{"key":"B142","doi-asserted-by":"publisher","first-page":"856","DOI":"10.3390\/math13050856","article-title":"Hallucination mitigation for retrieval-augmented large language models: a review","volume":"13","author":"Zhang","year":"2025","journal-title":"Mathematics"},{"key":"B143","doi-asserted-by":"publisher","first-page":"e72998","DOI":"10.2196\/72998","article-title":"Citation accuracy challenges posed by large language models","volume":"11","author":"Zhang","year":"2025","journal-title":"JMIR Med. Educ."},{"key":"B144","doi-asserted-by":"publisher","first-page":"239","DOI":"10.1038\/s41746-025-01651-w","article-title":"Leveraging long context in retrieval augmented language models for medical question answering","volume":"8","author":"Zhang","year":"2025","journal-title":"npj Digit. Med."},{"key":"B145","doi-asserted-by":"publisher","first-page":"14","DOI":"10.1038\/s44387-025-00019-5","article-title":"Exploring the role of large language models in the scientific method: from hypothesis to discovery","volume":"1","author":"Zhang","year":"2025","journal-title":"npj Artif. Intell."},{"key":"B146","article-title":"LLM hallucinations in the wild: large-scale evidence from non-existent citations","author":"Zhao","year":"2026"},{"key":"B147","doi-asserted-by":"publisher","first-page":"e57978","DOI":"10.2196\/57978","article-title":"The evaluation of generative AI should include repetition to assess stability","volume":"12","author":"Zhu","year":"2024","journal-title":"JMIR Mhealth Uhealth"},{"key":"B148","article-title":"LLM-based human-agent collaboration and interaction systems: a survey","author":"Zou","year":"2025"}],"container-title":["Frontiers in Toxicology"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/ftox.2026.1877128\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,28]],"date-time":"2026-09-28T02:07:55Z","timestamp":1790561275000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/ftox.2026.1877128\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,28]]},"references-count":148,"alternative-id":["10.3389\/ftox.2026.1877128"],"URL":"https:\/\/doi.org\/10.3389\/ftox.2026.1877128","relation":{},"ISSN":["2673-3080"],"issn-type":[{"value":"2673-3080","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9,28]]},"article-number":"1877128"}}