OpenAI00:00Press ReleasesOfficial Blog
OpenAI & PNNL aim to speed up permitting drafts
Speeds up citation-backed drafts for document-heavy work.
Key Points
- 1Built a benchmark for drafting NEPA sections
- 2Uses Codex CLI to work with a filesystem
- 3Potential 1–5 hours saved per subsection
- 4Experts scored structure, accuracy, and citations
- 5Production use still needs human review
OpenAI published results with PNNL on using AI to draft federal NEPA permitting sections faster. Using a coding agent (Codex CLI) could save about 1–5 hours per subsection. It’s a practical benchmark for doc-heavy work with citations. Humans must still review, especially for outdated or inconsistent sources.