Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Permit Ai topic
No spam. Unsubscribe anytime.
PNNL unveils Permit AI suite to speed NEPA reviews, offering searchable NEPA corpus and agency tools
Summary
Pacific Northwest National Laboratory demonstrated Permit AI, a suite of AI tools (SearchNEPA, CommentNEPA, WriteNEPA, EngageNEPA) built on a public NEPA Text Corpus to help federal subject‑matter experts find documents, process public comments, and draft sections of environmental reviews more quickly.
Pacific Northwest National Laboratory (PNNL) researchers described Permit AI on a webinar as a multi‑tool platform designed to accelerate federal environmental reviews conducted under the National Environmental Policy Act (NEPA).
Samira Horwala Pitana, a senior data scientist at PNNL and the project’s principal investigator, said Permit AI combines a curated NEPA text corpus with retrieval‑augmented models and four user‑facing applications: SearchNEPA/ChatNEPA for research and citation‑backed answers; CommentNEPA for ingesting and categorizing public comments; WriteNEPA for drafting template sections; and EngageNEPA for discovering public‑engagement notices. “We have seen less than 50% efficiency gain,” Samira said, describing early results for NEPA research tools.
The platform is grounded in a NEPA Text Corpus that PNNL reported contains more than 140,000 documents from more than 60 federal agencies. According to Samira, the corpus was expanded recently from several tens of thousands of documents and now represents billions of tokens, enabling the team to extract paragraph‑level metadata and supply citations in chat answers and drafting outputs.
Project team members said agencies and agency staff are the primary early users: more than 700 federal testers have tried the applications and SearchNEPA has drawn more than 200,000 page visits during beta testing. CommentNEPA, the public‑comment processing tool, was validated across more than 10 real‑world use cases (including DOE, the Nuclear Regulatory Commission and the Bureau of Land Management), and the team reported time savings in those validations (Samira characterized the average time savings in those validations as about 94%).
Samira emphasized that the tools are designed to assist subject‑matter experts rather than to replace legal or policy judgments. “These tools give citations so agency users can review where information is coming from,” she said, noting the team uses retrieval‑augmented generation (RAG) to ground outputs in source documents and lower hallucination risk.
PNNL and DOE staff also outlined partnership paths for deploying Permit AI at scale: technology transfer agreements to grant research or commercial rights, sponsored research for customized development, and collaborative research agreements for federally funded projects. Sarah (surname not specified) encouraged interested organizations to engage DOE’s Lab Partnering Service to explore licensing, CRADAs or sponsored research.
The presenters said most end‑user applications remain in federal beta testing and are not yet publicly available, though the NEPA Tech dataset (version 2) has been released on Hugging Face under PNNL’s account. The team said they plan further dataset releases and ongoing validation in collaboration with agency partners.
The webinar closed with an invitation to submit follow‑up questions by email, and presenters said they would continue outreach to agencies and prospective commercial or public partners.
