On Friday a ban on data centers becomes law in Coachella. Not a pause and not a study — a prohibition, written into the city’s code, passed by a council that took months of public scrutiny to get there.
It isn’t the only one. Palm Springs adopted a 45-day pause on the same day Coachella took its final vote. Indio and Desert Hot Springs each approved moratoriums of their own. Four cities in one summer, and none of them waited for Sacramento or Washington to tell them what to do.
I don’t think most people here know that happened.
It’s worth sitting with, because the question everyone keeps asking about AI — should it be regulated — is already settled. It is being regulated, quickly, in statehouses most of us never read about. The live question is who’s holding the pen.
Two days after California’s governor signed a law requiring ads that use AI-generated performers to say so, NPR published an investigation into who is actually writing the chatbot bills moving through the states. It is worth saying what that story is about, because I am about to use only one part of it. It is built around Cynthia Montoya, whose daughter Juliana was thirteen when she started talking to chatbots and died by suicide within a year. At least 75 lawsuits are pending against AI developers over harms to children. Those deaths are why the bills exist. What NPR found is who has been writing them. Several states, Hawaii and Iowa and Oregon and Washington among them, introduced chatbot bills this year largely modelled on a 2025 California law that consumer advocates already called industry-friendly. And lawmakers, lawyers and safety advocates told NPR that bills in at least ten states carried similar language. The exemptions are the tell. One carve-out, for a feature sitting inside another app, appears in bills in at least eight states. Another, in five states, exempts anything that works as a voice assistant — which would cover Alexa. NPR’s finding is that some of the bills already signed into law will exempt the most popular chatbots from the rules those bills were named for.
That’s not an argument against rules. It’s a warning about what rules look like when only one side shows up.
Which is why four councils in this valley matter more than their size suggests. They held the meetings, took the comment, and voted. That is the whole mechanism, and any city can use it.
A TV crew asked me on Monday what guardrails ought to look like, and I said transparency — I want to know when AI is put in front of me, whether that’s video, images or anything else. I still think that. It’s also the easy half.
There’s a second thing that didn’t come up on camera, and it worries me more than deepfakes do. Before I get to it, you should know who’s telling you. The survey work below is ours, published on aicoachellavalley.com, which is a product of my company, SunshineFM LLC. This newsroom is the nonprofit side. And AICV sells a program called Get Agent Ready that works on these gaps for money. Our own report puts it plainly: the program “operates at exactly the dimensions this audit measures”. The numbers are still the numbers. Weigh them knowing I have a business on the other side of them.
While we argue about whether to regulate these systems, most businesses here are hard for them to read. Two things are going on, and people run them together. I did it myself in an earlier draft of this.
One is a readiness problem. We went through 3,627 listings across the valley’s visitor economy and checked each against eight basic things software looks for. Does the site work. Do the listings agree with each other. Is there schema markup. That sort of thing. Fifty-eight got seven or eight of them right. That’s 1.6 percent. It doesn’t mean the rest can’t be read at all — most have a site that loads fine on a phone, 73 percent of them. It means almost nobody here has done the plumbing that makes a business easy for software to understand.
The other is simpler, and worse. It’s a door, and it’s either open or shut. We looked at 224 hotels and venues here and tried to reach the 148 with sites we could get to. Nearly a third turned an AI agent away at the door — 31.1 percent. Not incomplete. Blocked. And the walls are thickest around the big chains and the flagship resorts, which is exactly where a visitor with money to spend looks first.
None of it holds still, either. We went back a few days later and read 536 of them again, most of them restaurants rather than hotels. We could compare 490 against their first reading. About one in nine had already changed.
Those numbers point at the thing this whole conversation keeps missing. A label on AI content protects you as a consumer. It does nothing about whether your restaurant shows up when a visitor asks an assistant where to eat tonight. Nobody voted for that. It got decided by defaults, by companies, and by whoever writes the list.
Patrick Jacobs, an assistant professor of computer information systems at College of the Desert, told News Channel 3 that the point of going to the gym is stronger muscles, not lifted weights. Best line in the piece, and it’s true well past the classroom. The point of a rule isn’t having a rule. It’s ending up stronger than you started.
So here’s what I’d rather be asked than whether AI needs guardrails. Four councils here proved this summer that a city can read the reports, sit through the comment and write the rule itself. The next decision is already coming, because all this infrastructure is looking for somewhere to go. Will anyone from here be in the room — or will we read about it afterward?