AI policy is being decided by county commissions

Two years of argument about model licensing and compute thresholds, and the decisions actually constraining AI capacity are being taken by people weighing water use and noise ordinances.

Organised community opposition has increasingly stalled or cancelled US data centre projects, making municipal permitting the top practical constraint even where federal policy is supportive. The venue moved and most of the commentary did not follow.

An awkward battlefield

National advocacy can be professionalised — one regulator, one framework, one set of arguments. A thousand local hearings cannot. Each one can say no for reasons that have nothing to do with artificial intelligence, and community opposition does not need a position on frontier risk to reject a substation.

Money that hardens the opposition

The super PAC response is predictable and carries an obvious hazard: visible industry money aimed at local politics tends to confirm the suspicion it was meant to dispel. The trade actually on offer — speed in exchange for security and reliability commitments — gets struck at the permitting table, not in a legislature.

Meanwhile, the other venue

And where courts are deciding AI questions, they are doing it through doctrine written for other things. A hacking statute from another era now defines who is responsible when an agent acts. Neither of the places currently setting AI policy is the one anybody was watching.

If you want to know how much AI gets built next year, read planning agendas, not policy papers.

Brookings — Data center backlash signals a fight over AI power → · Institute for Progress — A speed-for-security bargain for AI data centers →