That’s what this is. Every week, EHS Risk Explained will cover one issue that EHS leaders need to understand. What is driving risk, where enforcement is headed, how AI and data are changing the work, and what it means for the people responsible for keeping their operations running and out of the wrong headlines.

This week’s issue is about AI. Specifically, about who’s already using it, and who should be.

The AI Enforcement Gap

Regulators can now run AI models to decide which facilities to inspect. Most EHS teams aren’t running that same analysis. That’s the gap.

What’s Changed

Inspection targeting has always involved judgment calls. An inspector’s experience, annual agency priorities, a complaint from a neighbor, a tip, a scheduled rotation. That’s how it worked for decades.

That’s still happening. But alongside it, government inspection programs are increasingly running predictive models to decide where to target inspections. They are pulling operational and compliance data through machine learning models to rank which facilities look like the best bets for violations. The gut call doesn’t go away. It gets a data layer on top of it.

This shifts enforcement from often reactive to targeted. Keeping a low profile used to reduce your odds. A facility that didn’t generate complaints, didn’t have a significant enforcement history, and kept its head down was less likely to see an inspector. AI-driven targeting models changed that. They flag on data, not reputation.

AI targeting for Inspections outperformed standard EPA targeting. In field testing from a UChicago/EPA machine learning model that began back in 2017, and has since expanded, inspections targeting RCRA hazardous waste facilities found violations at signficantly higher rates versus standard EPA targeting techniques. Same inspectors. Same rules. Different targeting. This type of model can now be applied across the regulatory spectrum.

What the Data Actually Shows

The data regulators are running through these models isn’t secret. Most of it is the same public compliance record your own EHS team already has access to. ECHO. OSHA’s inspection database. State agency records. The Toxic Release Inventory. National Response Center incident reports.

What they’re doing is normalizing it, weighting it, and running it against models trained on past violations. What you’ve emitted, how you’ve tracked against permit limits, how often you’ve had incidents, how your record compares to similar facilities. They’re building a picture of your site from the outside in.

If your team isn’t running the same kind of analysis on your own data, it’s worth asking why not.

Why Companies Get Caught Flat-Footed

In my experience, the companies that are targeted for inspection aren’t always the ones with the worst records. Often, they’re the ones whose EHS function never made the shift from filing paperwork to actually working its own data and applying it to assess and mitigate risk.

There is a version of this that’s been true for years. What’s new is the speed. AI compresses the time between a pattern showing up in public data and a regulator deciding your facility is worth a visit. The lag that used to give you lead time is smaller now.

And it’s not just regulators. NGOs have built data systems that would have sounded far-fetched five years ago. Your supply chain partners’ sustainability platforms can query the same databases on a rolling basis. The information head start that used to give companies lead time on compliance and reputational problems is mostly gone.

The Practical Question

I built Ecolumix because I came from the don’t-trust-verify world of data. Most of the tools out there run on voluntary reporting. We harvest the data companies already report to the government, across EPA, OSHA, state environmental agencies, the National Response Center, and more, and build facility-level profiles that are actually usable for screening and analysis.

The question for EHS teams isn’t whether AI is coming to enforcement. It’s already here. The question is whether your team is looking at your own facilities the way your regulators, your critics, and your business partners already are.

Until next week,

Doug Parker

CEO, Ecolumix

---

Coming in future issues:

Remote sensing comes of age. The equipment got cheap. Community groups figured that out and are building their own continuous emissions monitoring programs and their own records. That data doesn’t stay local.

Chemical incidents are on the rise. Accidents and releases are up, and federal regulators are paying attention. If chemicals move through your supply chain, not just your own operations, your exposure profile deserves a hard look.

Questions or topics worth covering? Reach out: doug@ecolumix.com