Virginia Just Put the Thing AI Runs On Under Review

Picture a resident at a county meeting asking what a proposed data center down the road will draw from the grid, and learning that the state agency involved signed an agreement not to say.

In Virginia, that answer just changed.

Governor Abigail Spanberger signed Executive Order 22 on September 18, creating a Data Center Accountability Framework. Her administration calls it "the most comprehensive and aggressive data center accountability effort in the country." The order bars executive branch agencies from signing nondisclosure agreements that hide material information about proposed data center projects. It ends state site-readiness and expedited permitting help for projects with anticipated peak demand of 25 megawatts or more. And it convenes a rapid-response AI Task Force to evaluate enforcement options under Virginia law, assess cybersecurity threats from AI agents and bots, and coordinate with frontier AI developers, not just data center operators.

Spanberger has already signaled where this goes next. She intends to seek 2027 legislation eliminating by-right approval for large data centers. A project that can now proceed once it meets zoning standards would instead need public hearings and an approval process.

Most AI governance fights so far have been fought on the same ground: what the model says, who it discriminates against, whether it hallucinated something that cost someone money or safety. Virginia just opened a second front, and almost no board has a policy for it yet.

It's not a technology order. It's an input order. The governance conversation everyone's been having is about what AI produces. This one is about what AI consumes, the energy, the water, the land, and whether a community gets a formal say before any of that gets committed.

The framework's own pillars read like the list of questions communities have been asking: transparency, environmental protection, energy affordability. And developers who assumed by-right approval was permanent now have to plan around the chance it isn't.

For a board outside Virginia, the mistake would be treating this as a regional zoning story. If your organization leases cloud capacity, co-locates workloads, or simply depends on a provider whose AI runs in a data center somewhere, the jurisdiction that data center sits in is now a variable in whether your service keeps running on the terms you signed up for.

This month, find out which state your AI infrastructure actually depends on. Then ask your vendor, in writing, what happens to your service if local approval in that state gets harder. Don't wait for your renewal cycle. Ask now, while the answer is still simple.

Read the full research behind this story, plus this week's senior housing and insurance coverage flagships, in the LITT Briefing on Substack.

If you want decisions like this on your board's agenda before they turn into a crisis, the ninety-day playbook is in my book, Senior Housing: The AI Gap: From Governance Theater to Deployment in 90 Days, available on Amazon. The compliance checklists and implementation tools are at youraigap.com.

← Back to Governance