EPA Lists 82 AI Use Cases but Deploys Only One High‑Impact Tool

*TL;DR: The EPA catalogues 82 AI projects but has only one high‑impact AI system actively deployed.

The agency’s 2025 AI Use Case Inventory, released in early 2026, enumerates 82 distinct AI applications across regulatory, administrative and support functions. The list combines deployed tools, pilots, pre‑deployment experiments and retired projects.

Only two items receive a high‑impact label: one already deployed and one still in pre‑deployment testing. A third entry is marked as “presumed high‑impact” but has not been formally assessed.

The deployed high‑impact system (item 78) uses machine learning to rank inspections of large‑quantity hazardous waste generators under the Resource Conservation and Recovery Act (RCRA). By analysing historic compliance data, the model directs inspectors to facilities most likely to violate regulations, cutting staff hours and improving violation detection.

Item 43, the pre‑deployment high‑impact case, targets the lead‑abatement program. It would scan documents, photos and videos to identify violations of the Toxic Substances Control Act’s lead‑disclosure rules. EPA officials say the tool could draft enforcement actions, but final decisions would remain a government function.

Item 46, labeled “presumed high‑impact,” is a surveillance‑camera analysis tool called Brief Cam. It flags relevant video segments for EPA Special Agents, streamlining evidence review.

Most of the remaining 79 entries are low‑level tasks such as scheduling, document comparison or help‑desk chat bots. These functions, while useful, do not alter regulatory outcomes.

Bridget C.E. Dooling, a policy analyst, argues that merely keeping a “human in the loop” does not satisfy the agency’s regulatory responsibilities. She urges that algorithms be fully integrated into decision‑making processes rather than serving as optional aides.

The disparity between the inventory’s breadth and actual deployment suggests that EPA’s AI ambitions outpace operational capacity. Federal guidance from the Office of Management and Budget requires agencies to produce annual AI inventories, but the inventory alone does not guarantee implementation.

What to watch next: EPA’s progress on the lead‑abatement AI pilot and any policy shifts that could move “presumed high‑impact” tools into active use.