AI cuts chemical risk assessment time by over 70%

South Korea's National Institute of Environmental Research, under the Ministry of Climate, Environment and Energy, announced Wednesday that it has developed an AI-powered technology that automatically collects and extracts hazard data used in chemical risk assessments and compiles it into a database — and has filed an international patent application for the system.

Risk assessment is the process of determining the degree of danger posed by a chemical substance by combining an evaluation of how harmful it is with an analysis of how much it affects people and the environment. The process follows four sequential steps: hazard identification, dose-response assessment, exposure assessment and risk characterization.

Until now, experts had to manually search through and review vast volumes of domestic and foreign materials at each stage, limiting the number of substances that could be assessed in a given year.

The newly developed technology automates the first step — hazard identification — using AI across three stages: surveying and collecting domestic and foreign reports, automatically extracting hazard data, and building a database.

Assessors can also immediately cross-check AI-extracted data against the original source documents, which the institute said enhances the reliability and validity of hazardous chemical evaluations.

The institute has developed a pilot model called ATOM that applies this technology and is already using it to collect and extract hazard data from key sources, including EU risk assessment reports.

The goal is to progressively automate the entire process — from hazard data collection through exposure assessment, risk characterization and report writing — by 2030, cutting the time required to assess each substance by more than 70 percent compared with current timelines.

Building on these AI-driven automation results, the institute plans to work with governments of 10 countries including the Netherlands, as well as the UN, OECD and EU, to raise the standard of risk assessment technology.

"This international patent filing lays the groundwork for rapidly and consistently deriving information automatically from vast hazard datasets," said Kim Su-jin, head of the institute's Environmental Health Research Division. "It will allow us to expand the number of substances assessed annually — currently around 15 — to more than three times that figure."