RICHLAND, Wash.-An innovative public-private partnership has yielded powerful new tools to help federal agencies rapidly synthesize complex data, historical documents and more into draft environmental impact statements.
The benchmark project, called DraftNEPABench, paired the Department of Energy's Pacific Northwest National Laboratory with corporate partner OpenAI. PNNL's PermitAITM research team worked together with OpenAI over the past year to evaluate whether AI agents originally designed for coding tasks could help draft complex sections of environmental impact statements. These major reports are often required under the National Environmental Policy Act (NEPA) when agencies consider major federal projects such as siting of new data centers or electricity infrastructure.
"Our evaluation showed that AI coding agents can generate structured and domain-specific draft sections for environmental impact statements with promising results," said PNNL data scientist and DraftNEPABench research lead Anurag Acharya. "While the systems still require human oversight, the benchmark highlights both the potential and current limitations of these approaches."
The research team presented the peer-reviewed research at the first annual Association of Computing Machinery Conference on AI and Agentic Systems, held May 27-29, 2026, in San Jose, Calif.
PermitAITM began as a pilot project sponsored by the DOE's Office of Policy to centralize NEPA decision data. Driven by the recent federal priority to accelerate and improve environmental reviews, the PermitAITM team has been steadily expanding its focus.
What makes DraftNEPABench unique?
Benchmarks are like standardized tests for AI systems. Researchers use them to evaluate coding and scientific algorithms. But rigorous benchmarks for drafting documents in regulatory settings are largely absent. That's because until the PermitAITM team developed data standards and metadata coding, most environmental permitting documentation remained siloed and inaccessible. Once the team released the machine-readable dataset, known as NEPA Text Corpus (NEPATEC), accessing historical NEPA data and decisions made labor-intensive searches much simpler.
