AI expert and industry leading toxicologist Thomas Hartung hails launch of agentic AI platform a “transformative moment” in chemical safety science

(Press-News.org) BALTIMORE, MD, March 14, 2026, Dr. Thomas Hartung, Director of the Center for Alternatives to Animal Testing (CAAT) at the Johns Hopkins Bloomberg School of Public Health, has endorsed the public launch of an agentic AI platform developed by Insilica Inc. that produces comprehensive, source-traceable toxicological risk assessments in just a few hours.

The launch of ToxIndex meets a critical need in chemical and drug safety as well as within field of exposomics, a field of study which considers environmental exposures effect on human health and serves as a compliment and counterpart to genomics. An agentic platform with the ability to autonomously source traceable exposure and toxicity datasets has the potential to drastically improve data harmonization within the field and across relevant subdisciplines. For most fields, data harmonization exists as a stopgap to maximal research progress and discovery potential. For exposomics, this is especially true given the fields expansive scope and interest in the entire life course as opposed to singular diagnoses or exposures. Automated and tailored access to historical studies and relevant data sources will almost certainly contribute to new and improved inferences by integrating historical data sources and advancing capacities within the field.

Two Decades Toward a New Toxicology

Safety testing of chemicals costs close to $20 billion per year, but this has sufficed to assess only about 1 in 10 chemicals in commercial use; 1 in 100 receives a comprehensive evaluation. The 2007 National Research Council report Toxicity Testing in the 21st Century articulated a vision for replacing traditional animal-based methods with integrated computational models, in vitro assays, and pathway-based approaches. In the nearly two decades since, the scientific community has built an extraordinary portfolio of New Approach Methodologies (NAMs): validated QSAR models, high-throughput screening libraries like ToxCast and Tox21, adverse outcome pathway frameworks, and curated regulatory databases.

ToxIndex: The Integration Layer this Field and Others Have Long Been Missing

The new platform addresses this gap directly. Developed by Thomas Luechtefeld, who completed his doctoral work in computational toxicology at CAAT under Dr. Hartung’s supervision, the platform uses AI agents that programmatically access and orchestrate the full landscape of toxicological resources.

The platform integrates three tiers of New Approach Methodologies into a single, auditable workflow. In silico predictions draw on over 600 containerized open-source toxicology models and a proprietary transformer trained on 254 million human chemical activity measurements. In vitro evidence is indexed from a knowledge graph of over 60 billion data triples, co-developed with NICEATM under NSF Proto-OKN funding, encompassing ToxCast, Tox21, ChEMBL, PubChem, and 1,200 additional curated sources. Physicochemical and ADME profiling rounds out the evidence base with data retrieved from REACH dossiers, EPA databases, and computational tools like the OECD QSAR Toolbox.

Every claim in a ToxIndex-generated assessment carries full provenance: traceable to the specific source database, table, row, and column of origin. This level of traceability meets and in many cases exceeds current REACH and TSCA standards for regulatory documentation.

In a proof-of-concept study, ToxIndex generated a 47-page, 944-claim toxicological risk assessment for dodecanedioic acid (DDDA) in under three hours. This is a task that conventionally requires three to six months of expert manual work. Every claim carries a Klimisch reliability score and is traceable to its source. The report updates automatically as new literature is indexed by the platform’s document crawler, which continuously processes regulatory and scientific texts from agencies worldwide.

ToxIndex can also re-evaluate as new data arrives, route procurement requests to testing laboratories when evidence gaps remain and integrate every new NAM into a shared knowledge graph where each contribution compounds the value of every other.

A Research Partnership Spanning Continents

Dr. Hartung and Dr. Luechtefeld also collaborate within the European ONTOX consortium, a €17.2 million EU Horizon 2020 initiative developing non-animal toxicity prediction methods. Through ToxTrack LLC, Insilica’s sister company, the platform has already been integrated with European partners including EU regulatory scientists, validating the technology against international requirements.

The collaboration will now be extended to broad public use, through sharing of concepts on arXiv, code on GitHub and making the underlying Transformer freely available – discussions with NIH and OECD are on the way.  Thereby, the platform is being positioned as the integration infrastructure that makes every participating NAM permanently useful. As novel methods such as complex in vitro models, in silico systems, and in chemico approaches become available, ToxIndex’s agents automatically discover and register them, linking each new method to the regulatory endpoints and chemical domains it covers.

A Regulatory Inflection Point

The timing of ToxIndex’s launch reflects converging regulatory mandates that make AI-driven safety assessment not merely attractive but necessary. The FDA Modernization Act 2.0 has eliminated the requirement for mandatory animal testing and explicitly endorsed computational methods as alternatives. The EPA has committed to eliminating mammalian testing by 2035. Europe’s €9.5 billion REACH compliance burden has created massive backlogs, while the EU’s animal testing ban for cosmetics has made AI-first safety assessment a strategic imperative for one of the world’s largest consumer products industries.

Meanwhile, AI-driven drug discovery is generating thousands of novel molecules that each require safety evaluation. Every chemical designed by AI needs a safety assessment, yet only 9,000 toxicologists serve all of North America, and 88% of new chemical reviews already exceed statutory deadlines.

Dr. Hartung’s seven years with the European Commission as head of ECVAM (the European Centre for the Validation of Alternative Methods), combined with his advisory roles with the EPA, FDA, EFSA, OECD, and Apple’s Green Chemistry Advisory Board, provide the regulatory trust network that underpins Insilica’s credibility with agencies and industry alike. His endorsement carries particular weight in a field where, as he notes, 95% accuracy is not good enough when human lives are at stake: “Trust must be earned through institutional relationships, transparent methodology, and rigorous validation.”

Statements

“We wrote the scientific papers. We built the databases. We validated the models. We advocated for policy change. What we lacked was the engineering to make it all work together at the scale regulators and industry actually need. Tom Luechtefeld has built that engineering. ToxIndex brings the 2007 NRC vision to life, not as a research prototype but as production infrastructure that is already delivering regulatory-grade assessments to the FDA, Fortune 500 companies, and European regulators. I have been in this field for over thirty years, and I can say without hesitation: this is a transformative moment.”

— Dr. Thomas Hartung, Director, Center for Alternatives to Animal Testing (CAAT), Johns Hopkins Bloomberg School of Public Health; former Head of ECVAM, European Commission; Scientific Advisor, Insilica Inc.

“Everything I know about toxicology, I learned at CAAT. Professor Hartung taught me that 21st-century toxicology was not about building one better model, it was about building the system that makes every model, every database, and every assay platform work together. ToxIndex is that system. It is deeply gratifying to launch it at the same conference where this community gathers every year to push the science forward.”

— Thomas Luechtefeld, PhD, Founder & CEO, Insilica Inc.; CAAT alumnus (PhD, computational toxicology), Johns Hopkins part-time Research Associate

At the Society of Toxicology Annual Meeting

Insilica will present and exhibit at the SOT 65th Annual Meeting at the San Diego Convention Center, March 22–25, 2026. Live demonstrations of the new tool will include real-time agentic risk assessments, IUCLID dossier generation for REACH compliance, and interactive exploration of the Brickyard knowledge graph. Dr. Hartung and Dr. Luechtefeld will be available for meetings and media interviews throughout the conference.

About Insilica Inc.

Insilica is the full-service AI toxicology company. Combining proprietary AI, expert toxicologists, and the world’s largest chemical safety data infrastructure, Insilica delivers turnkey safety assessments that empower CROs, consultancies, and internal teams to scale beyond what traditional approaches allow. Founded in 2017 by CAAT alumnus Thomas Luechtefeld, the company is a direct contractor to the FDA, a participant in the EU ONTOX consortium, and serves enterprise customers across pharma, energy, agriculture, cosmetics, and consumer products. Insilica is based in Rockville, Maryland.

Website: insilica.co  |  Platform: sage.toxindex.com

About the Center for Alternatives to Animal Testing (CAAT)

The Johns Hopkins Center for Alternatives to Animal Testing (CAAT) is one of the world’s leading academic centers dedicated to the development and validation of non-animal methods for assessing the safety of chemicals, drugs, and consumer products. Founded in 1981, CAAT promotes humane science by refining, reducing, and replacing the use of animals in research, product safety testing, and education. CAAT also serves as the U.S. headquarters for the Transatlantic Think Tank for Toxicology (t4), a collaboration with CAAT-Europe at the University of Konstanz. Under the direction of Dr. Thomas Hartung, CAAT has been at the forefront of evidence-based toxicology and the development of computational approaches to chemical safety evaluation.

Website: caat.jhsph.edu

END